arXiv Daily Digest - 2026-03-06

Scaling imitation learning is fundamentally constrained by the efficiency of data collection. While handheld interfaces have emerged as a scalable solution for in-the-wild data acquisition, they predominantly operate in an open-loop manner: operators blindly collect demonstrations without knowing the underlying policy's weaknesses, leading to inefficient coverage of critical state distributions. Conversely, interactive methods like DAgger effectively address covariate shift but rely on physical robot execution, which is costly and difficult to scale. To reconcile this trade-off, we introduce RoboPocket, a portable system that enables Robot-Free Instant Policy Iteration using single consumer smartphones. Its core innovation is a Remote Inference framework that visualizes the policy's predicted trajectory via Augmented Reality (AR) Visual Foresight. This immersive feedback allows collectors to proactively identify potential failures and focus data collection on the policy's weak regions without requiring a physical robot. Furthermore, we implement an asynchronous Online Finetuning pipeline that continuously updates the policy with incoming data, effectively closing the learning loop in minutes. Extensive experiments demonstrate that RoboPocket adheres to data scaling laws and doubles the data efficiency compared to offline scaling strategies, overcoming their long-standing efficiency bottleneck. Moreover, our instant iteration loop also boosts sample efficiency by up to 2$\times$ in distributed environments a small number of interactive corrections per person. Project page and videos: https://robo-pocket.github.io.

Efficient and stable training of large language models (LLMs) remains a core challenge in modern machine learning systems. To address this challenge, Reparameterized Orthogonal Equivalence Training (POET), a spectrum-preserving framework that optimizes each weight matrix through orthogonal equivalence transformation, has been proposed. Although POET provides strong training stability, its original implementation incurs high memory consumption and computational overhead due to intensive matrix multiplications. To overcome these limitations, we introduce POET-X, a scalable and memory-efficient variant that performs orthogonal equivalence transformations with significantly reduced computational cost. POET-X maintains the generalization and stability benefits of POET while achieving substantial improvements in throughput and memory efficiency. In our experiments, POET-X enables the pretraining of billion-parameter LLMs on a single Nvidia H100 GPU, and in contrast, standard optimizers such as AdamW run out of memory under the same settings.

We study two recurring phenomena in Transformer language models: massive activations, in which a small number of tokens exhibit extreme outliers in a few channels, and attention sinks, in which certain tokens attract disproportionate attention mass regardless of semantic relevance. Prior work observes that these phenomena frequently co-occur and often involve the same tokens, but their functional roles and causal relationship remain unclear. Through systematic experiments, we show that the co-occurrence is largely an architectural artifact of modern Transformer design, and that the two phenomena serve related but distinct functions. Massive activations operate globally: they induce near-constant hidden representations that persist across layers, effectively functioning as implicit parameters of the model. Attention sinks operate locally: they modulate attention outputs across heads and bias individual heads toward short-range dependencies. We identify the pre-norm configuration as the key choice that enables the co-occurrence, and show that ablating it causes the two phenomena to decouple.

To scale the solution of optimization and simulation problems, prior work has explored machine-learning surrogates that inexpensively map problem parameters to corresponding solutions. Commonly used approaches, including supervised and self-supervised learning with either soft or hard feasibility enforcement, face inherent challenges such as reliance on expensive, high-quality labels or difficult optimization landscapes. To address their trade-offs, we propose a novel framework that first collects "cheap" imperfect labels, then performs supervised pretraining, and finally refines the model through self-supervised learning to improve overall performance. Our theoretical analysis and merit-based criterion show that labeled data need only place the model within a basin of attraction, confirming that only modest numbers of inexact labels and training epochs are required. We empirically validate our simple three-stage strategy across challenging domains, including nonconvex constrained optimization, power-grid operation, and stiff dynamical systems, and show that it yields faster convergence; improved accuracy, feasibility, and optimality; and up to 59x reductions in total offline cost.

Large language models sometimes produce false or misleading responses. Two approaches to this problem are honesty elicitation -- modifying prompts or weights so that the model answers truthfully -- and lie detection -- classifying whether a given response is false. Prior work evaluates such methods on models specifically trained to lie or conceal information, but these artificial constructions may not resemble naturally-occurring dishonesty. We instead study open-weights LLMs from Chinese developers, which are trained to censor politically sensitive topics: Qwen3 models frequently produce falsehoods about subjects like Falun Gong or the Tiananmen protests while occasionally answering correctly, indicating they possess knowledge they are trained to suppress. Using this as a testbed, we evaluate a suite of elicitation and lie detection techniques. For honesty elicitation, sampling without a chat template, few-shot prompting, and fine-tuning on generic honesty data most reliably increase truthful responses. For lie detection, prompting the censored model to classify its own responses performs near an uncensored-model upper bound, and linear probes trained on unrelated data offer a cheaper alternative. The strongest honesty elicitation techniques also transfer to frontier open-weights models including DeepSeek R1. Notably, no technique fully eliminates false responses. We release all prompts, code, and transcripts.

The growing complexity of hardware design and the widening gap between high-level specifications and register-transfer level (RTL) implementation hinder rapid prototyping and system design. We introduce NL2GDS (Natural Language to Layout), a novel framework that leverages large language models (LLMs) to translate natural language hardware descriptions into synthesizable RTL and complete GDSII layouts via the open-source OpenLane ASIC flow. NL2GDS employs a modular pipeline that captures informal design intent, generates HDL using multiple LLM engines and verifies them, and orchestrates automated synthesis and layout. Evaluations on ISCAS'85 and ISCAS'89 benchmark designs demonstrate up to 36% area reduction, 35% delay reduction, and 70% power savings compared to baseline designs, highlighting its potential to democratize ASIC design and accelerate hardware innovation.

We provide evidence of performative chain-of-thought (CoT) in reasoning models, where a model becomes strongly confident in its final answer, but continues generating tokens without revealing its internal belief. Our analysis compares activation probing, early forced answering, and a CoT monitor across two large models (DeepSeek-R1 671B & GPT-OSS 120B) and find task difficulty-specific differences: The model's final answer is decodable from activations far earlier in CoT than a monitor is able to say, especially for easy recall-based MMLU questions. We contrast this with genuine reasoning in difficult multihop GPQA-Diamond questions. Despite this, inflection points (e.g., backtracking, 'aha' moments) occur almost exclusively in responses where probes show large belief shifts, suggesting these behaviors track genuine uncertainty rather than learned "reasoning theater." Finally, probe-guided early exit reduces tokens by up to 80% on MMLU and 30% on GPQA-Diamond with similar accuracy, positioning attention probing as an efficient tool for detecting performative reasoning and enabling adaptive computation.

As AI models progress beyond simple chatbots into more complex workflows, we draw ever closer to the event horizon beyond which AI systems will be utilized in autonomous, self-maintaining feedback loops. Any autonomous AI system will depend on automated, verifiable rewards and feedback; in settings where ground truth is sparse or non-deterministic, one practical source of such rewards is an LLM-as-a-Judge. Although LLM judges continue to improve, the literature has yet to introduce systems capable of enforcing standards with strong guarantees, particularly when bias vectors are unknown or adversarially discovered. To remedy this issue, we propose average bias-boundedness (A-BB), an algorithmic framework which formally guarantees reductions of harm/impact as a result of any measurable bias in an LLM judge. Evaluating on Arena-Hard-Auto with four LLM judges, we achieve (tau=0.5, delta=0.01) bias-bounded guarantees while retaining 61-99% correlation with original rankings across formatting and schematic bias settings, with most judge-bias combinations exceeding 80%. The code to reproduce our findings is available at https://github.com/penfever/bias-bounded-evaluation.

Estimating heterogeneous treatment effects (HTEs) from right-censored survival data is critical in high-stakes applications such as precision medicine and individualized policy-making. Yet, the survival analysis setting poses unique challenges for HTE estimation due to censoring, unobserved counterfactuals, and complex identification assumptions. Despite recent advances, from Causal Survival Forests to survival meta-learners and outcome imputation approaches, evaluation practices remain fragmented and inconsistent. We introduce SurvHTE-Bench, the first comprehensive benchmark for HTE estimation with censored outcomes. The benchmark spans (i) a modular suite of synthetic datasets with known ground truth, systematically varying causal assumptions and survival dynamics, (ii) semi-synthetic datasets that pair real-world covariates with simulated treatments and outcomes, and (iii) real-world datasets from a twin study (with known ground truth) and from an HIV clinical trial. Across synthetic, semi-synthetic, and real-world settings, we provide the first rigorous comparison of survival HTE methods under diverse conditions and realistic assumption violations. SurvHTE-Bench establishes a foundation for fair, reproducible, and extensible evaluation of causal survival methods. The data and code of our benchmark are available at: https://github.com/Shahriarnz14/SurvHTE-Bench .

Singular statistical models-including mixtures, matrix factorization, and neural networks-violate regular asymptotics due to parameter non-identifiability and degenerate Fisher geometry. Although singular learning theory characterizes marginal likelihood behavior through invariants such as the real log canonical threshold and singular fluctuation, these quantities remain difficult to interpret operationally. At the same time, widely used criteria such as WAIC and WBIC appear disconnected from underlying singular geometry. We show that posterior tempering induces a one-parameter deformation of the posterior distribution whose associated observables generate a hierarchy of thermodynamic response functions. A universal covariance identity links derivatives of tempered expectations to posterior fluctuations, placing WAIC, WBIC, and singular fluctuation within a unified response framework. Within this framework, classical quantities from singular learning theory acquire natural thermodynamic interpretations: RLCT governs the leading free-energy slope, singular fluctuation corresponds to curvature of the tempered free energy, and WAIC measures predictive fluctuation. We formalize an observable algebra that quotients out non-identifiable directions, allowing structurally meaningful order parameters to be constructed in singular models. Across canonical singular examples-including symmetric Gaussian mixtures, reduced-rank regression, and overparameterized neural networks-we empirically demonstrate phase-transition-like behavior under tempering. Order parameters collapse, susceptibilities peak, and complexity measures align with structural reorganization in posterior geometry. Our results suggest that thermodynamic response theory provides a natural organizing framework for interpreting complexity, predictive variability, and structural reorganization in singular Bayesian learning.

Trustworthiness is a core research challenge for agentic AI systems built on Large Language Models (LLMs). To enhance trust, natural language claims from diverse sources, including human-written text, web content, and model outputs, are commonly checked for factuality by retrieving external knowledge and using an LLM to verify the faithfulness of claims to the retrieved evidence. As a result, such methods are constrained by retrieval errors and external data availability, while leaving the models intrinsic fact-verification capabilities largely unused. We propose the task of fact-checking without retrieval, focusing on the verification of arbitrary natural language claims, independent of their source. To study this setting, we introduce a comprehensive evaluation framework focused on generalization, testing robustness to (i) long-tail knowledge, (ii) variation in claim sources, (iii) multilinguality, and (iv) long-form generation. Across 9 datasets, 18 methods and 3 models, our experiments indicate that logit-based approaches often underperform compared to those that leverage internal model representations. Building on this finding, we introduce INTRA, a method that exploits interactions between internal representations and achieves state-of-the-art performance with strong generalization. More broadly, our work establishes fact-checking without retrieval as a promising research direction that can complement retrieval-based frameworks, improve scalability, and enable the use of such systems as reward signals during training or as components integrated into the generation process.

Real-time reconstruction of conditional quantum states from continuous measurement records is a fundamental requirement for quantum feedback control, yet standard stochastic master equation (SME) solvers require exact model specification, known system parameters, and are sensitive to parameter mismatch. While neural sequence models can fit these stochastic dynamics, the unconstrained predictors can violate physicality such as positivity or trace constraints, leading to unstable rollouts and unphysical estimates. We propose a Kraus-structured output layer that converts the hidden representation of a generic sequence backbone into a completely positive trace preserving (CPTP) quantum operation, yielding physically valid state updates by construction. We instantiate this layer across diverse backbones, RNN, GRU, LSTM, TCN, ESN and Mamba; including Neural ODE as a comparative baseline, on stochastic trajectories characterized by parameter drift. Our evaluation reveals distinct trade-offs between gating mechanisms, linear recurrence, and global attention. Across all models, Kraus-LSTM achieves the strongest results, improving state estimation quality by 7% over its unconstrained counterpart while guaranteeing physically valid predictions in non-stationary regimes.

Reading comprehension systems for low-resource languages face significant challenges in handling unanswerable questions. These systems tend to produce unreliable responses when correct answers are absent from context. To solve this problem, we introduce NCTB-QA, a large-scale Bangla question answering dataset comprising 87,805 question-answer pairs extracted from 50 textbooks published by Bangladesh's National Curriculum and Textbook Board. Unlike existing Bangla datasets, NCTB-QA maintains a balanced distribution of answerable (57.25%) and unanswerable (42.75%) questions. NCTB-QA also includes adversarially designed instances containing plausible distractors. We benchmark three transformer-based models (BERT, RoBERTa, ELECTRA) and demonstrate substantial improvements through fine-tuning. BERT achieves 313% relative improvement in F1 score (0.150 to 0.620). Semantic answer quality measured by BERTScore also increases significantly across all models. Our results establish NCTB-QA as a challenging benchmark for Bangla educational question answering. This study demonstrates that domain-specific fine-tuning is critical for robust performance in low-resource settings.

The process of debating is essential in our daily lives, whether in studying, work activities, simple everyday discussions, political debates on TV, or online discussions on social networks. The range of uses for debates is broad. Due to the diverse applications, structures, and formats of debates, developing corpora that account for these variations can be challenging, and the scarcity of debate corpora in the state of the art is notable. For this reason, the current research proposes the DEBISS corpus: a collection of spoken and individual debates with semi-structured features. With a broad range of NLP task annotations, such as speech-to-text, speaker diarization, argument mining, and debater quality assessment.

Attention, as a core layer of the ubiquitous Transformer architecture, is the bottleneck for large language models and long-context applications. While FlashAttention-3 optimized attention for Hopper GPUs through asynchronous execution and warp specialization, it primarily targets the H100 architecture. The AI industry has rapidly transitioned to deploying Blackwell-based systems such as the B200 and GB200, which exhibit fundamentally different performance characteristics due to asymmetric hardware scaling: tensor core throughput doubles while other functional units (shared memory bandwidth, exponential units) scale more slowly or remain unchanged. We develop several techniques to address these shifting bottlenecks on Blackwell GPUs: (1) redesigned pipelines that exploit fully asynchronous MMA operations and larger tile sizes, (2) software-emulated exponential and conditional softmax rescaling that reduces non-matmul operations, and (3) leveraging tensor memory and the 2-CTA MMA mode to reduce shared memory traffic and atomic adds in the backward pass. We demonstrate that our method, FlashAttention-4, achieves up to 1.3$\times$ speedup over cuDNN 9.13 and 2.7$\times$ over Triton on B200 GPUs with BF16, reaching up to 1613 TFLOPs/s (71% utilization). Beyond algorithmic innovations, we implement FlashAttention-4 entirely in CuTe-DSL embedded in Python, achieving 20-30$\times$ faster compile times compared to traditional C++ template-based approaches while maintaining full expressivity.

Establishing common ground, a shared set of beliefs and mutually recognized facts, is fundamental to collaboration, yet remains a challenge for current AI systems, especially in multimodal, multiparty settings, where the collaborators bring different information to the table. We introduce the Distributed Partial Information Puzzle (DPIP), a collaborative construction task that elicits rich multimodal communication under epistemic asymmetry. We present a multimodal dataset of these interactions, annotated and temporally aligned across speech, gesture, and action modalities to support reasoning over propositional content and belief dynamics. We then evaluate two paradigms for modeling common ground (CG): (1) state-of-the-art large language models (LLMs), prompted to infer shared beliefs from multimodal updates, and (2) an axiomatic pipeline grounded in Dynamic Epistemic Logic (DEL) that incrementally performs the same task. Results on the annotated DPIP data indicate that it poses a challenge to modern LLMs' abilities to track both task progression and belief state.

Current video generation models cannot simulate physical consequences of 3D actions like forces and robotic manipulations, as they lack structural understanding of how actions affect 3D scenes. We present RealWonder, the first real-time system for action-conditioned video generation from a single image. Our key insight is using physics simulation as an intermediate bridge: instead of directly encoding continuous actions, we translate them through physics simulation into visual representations (optical flow and RGB) that video models can process. RealWonder integrates three components: 3D reconstruction from single images, physics simulation, and a distilled video generator requiring only 4 diffusion steps. Our system achieves 13.2 FPS at 480x832 resolution, enabling interactive exploration of forces, robot actions, and camera controls on rigid objects, deformable bodies, fluids, and granular materials. We envision RealWonder opens new opportunities to apply video models in immersive experiences, AR/VR, and robot learning. Our code and model weights are publicly available in our project website: https://liuwei283.github.io/RealWonder/

Contact-rich micromanipulation in microfluidic flow is challenging because small disturbances can break pushing contact and induce large lateral drift. We study planar cell pushing with a magnetic rolling microrobot that tracks a waypoint-sampled reference curve under time-varying Poiseuille flow. We propose a hybrid controller that augments a nominal MPC with a learned residual policy trained by SAC. The policy outputs a bounded 2D velocity correction that is contact-gated, so residual actions are applied only during robot--cell contact, preserving reliable approach behavior and stabilizing learning. All methods share the same actuation interface and speed envelope for fair comparisons. Experiments show improved robustness and tracking accuracy over pure MPC and PID under nonstationary flow, with generalization from a clover training curve to unseen circle and square trajectories. A residual-bound sweep identifies an intermediate correction limit as the best trade-off, which we use in all benchmarks.

Imitation Learning (IL) enables agents to mimic expert behavior by learning from demonstrations. However, traditional IL methods require large amounts of medium-to-high-quality demonstrations as well as actions of expert demonstrations, both of which are often unavailable. To reduce this need, we propose Latent Wasserstein Adversarial Imitation Learning (LWAIL), a novel adversarial imitation learning framework that focuses on state-only distribution matching. It benefits from the Wasserstein distance computed in a dynamics-aware latent space. This dynamics-aware latent space differs from prior work and is obtained via a pre-training stage, where we train the Intention Conditioned Value Function (ICVF) to capture a dynamics-aware structure of the state space using a small set of randomly generated state-only data. We show that this enhances the policy's understanding of state transitions, enabling the learning process to use only one or a few state-only expert episodes to achieve expert-level performance. Through experiments on multiple MuJoCo environments, we demonstrate that our method outperforms prior Wasserstein-based IL methods and prior adversarial IL methods, achieving better results across various tasks.

World models provide a powerful framework for simulating environment dynamics conditioned on actions or instructions, enabling downstream tasks such as action planning or policy learning. Recent approaches leverage world models as learned simulators, but its application to decision-time planning remains computationally prohibitive for real-time control. A key bottleneck lies in latent representations: conventional tokenizers encode each observation into hundreds of tokens, making planning both slow and resource-intensive. To address this, we propose CompACT, a discrete tokenizer that compresses each observation into as few as 8 tokens, drastically reducing computational cost while preserving essential information for planning. An action-conditioned world model that occupies CompACT tokenizer achieves competitive planning performance with orders-of-magnitude faster planning, offering a practical step toward real-world deployment of world models.

Weakly-Supervised Dense Video Captioning aims to localize and describe events in videos trained only on caption annotations, without temporal boundaries. Prior work introduced an implicit supervision paradigm based on Gaussian masking and complementary captioning. However, existing method focuses merely on generating non-overlapping masks without considering their semantic relationship to corresponding events, resulting in simplistic, uniformly distributed masks that fail to capture semantically meaningful regions. Moreover, relying solely on ground-truth captions leads to sub-optimal performance due to the inherent sparsity of existing datasets. In this work, we propose SAIL, which constructs semantically-aware masks through cross-modal alignment. Our similarity aware training objective guides masks to emphasize video regions with high similarity to their corresponding event captions. Furthermore, to guide more accurate mask generation under sparse annotation settings, we introduce an LLM-based augmentation strategy that generates synthetic captions to provide additional alignment signals. These synthetic captions are incorporated through an inter-mask mechanism, providing auxiliary guidance for precise temporal localization without degrading the main objective. Experiments on ActivityNet Captions and YouCook2 demonstrate state-of-the-art performance on both captioning and localization metrics.

Reasoning models think out loud, but much of what they say is noise. We introduce OPSDC (On-Policy Self-Distillation for Reasoning Compression), a method that teaches models to reason more concisely by distilling their own concise behavior back into themselves. The entire approach reduces to one idea: condition the same model on a "be concise" instruction to obtain teacher logits, and minimize per-token reverse KL on the student's own rollouts. No ground-truth answers, no token budgets, no difficulty estimators. Just self-distillation. Yet this simplicity belies surprising sophistication: OPSDC automatically compresses easy problems aggressively while preserving the deliberation needed for hard ones. On Qwen3-8B and Qwen3-14B, we achieve 57-59% token reduction on MATH-500 while improving accuracy by 9-16 points absolute. On AIME 2024, the 14B model gains 10 points with 41% compression. The secret? Much of what reasoning models produce is not just redundant-it is actively harmful, compounding errors with every unnecessary token.

Practitioners have access to an abundance of language models and prompting strategies for solving many language modeling tasks; yet prior work shows that modeling performance is highly sensitive to both choices. Classical machine learning ensembling techniques offer a principled approach: aggregate predictions from multiple sources to achieve better performance than any single one. However, applying ensembling to language models during decoding is challenging: naively aggregating next-token probabilities yields samples from a locally normalized, biased approximation of the generally intractable ensemble distribution over strings. In this work, we introduce a unified framework for composing $K$ language models into $f$-ensemble distributions for a wide range of functions $f\colon\mathbb{R}_{\geq 0}^{K}\to\mathbb{R}_{\geq 0}$. To sample from these distributions, we propose a byte-level sequential Monte Carlo (SMC) algorithm that operates in a shared character space, enabling ensembles of models with mismatching vocabularies and consistent sampling in the limit. We evaluate a family of $f$-ensembles across prompt and model combinations for various structured text generation tasks, highlighting the benefits of alternative aggregation strategies over traditional probability averaging, and showing that better posterior approximations can yield better ensemble performance.

Image-to-3D generation faces inherent semantic ambiguity under occlusion, where partial observation alone is often insufficient to determine object category. In this work, we formalize text-driven amodal 3D generation, where text prompts steer the completion of unseen regions while strictly preserving input observation. Crucially, we identify that these objectives demand distinct control granularities: rigid control for the observation versus relaxed structural control for the prompt. To this end, we propose RelaxFlow, a training-free dual-branch framework that decouples control granularity via a Multi-Prior Consensus Module and a Relaxation Mechanism. Theoretically, we prove that our relaxation is equivalent to applying a low-pass filter on the generative vector field, which suppresses high-frequency instance details to isolate geometric structure that accommodates the observation. To facilitate evaluation, we introduce two diagnostic benchmarks, ExtremeOcc-3D and AmbiSem-3D. Extensive experiments demonstrate that RelaxFlow successfully steers the generation of unseen regions to match the prompt intent without compromising visual fidelity.

The ability to interpret machine learning model decisions is critical in such domains as healthcare, where trust in model predictions is as important as their accuracy. Inspired by the development of prototype parts-based deep neural networks in computer vision, we propose a new model for tabular data, specifically tailored to medical records, that requires discretization of diagnostic result norms. Unlike the original vision models that rely on the spatial structure, our method employs trainable patching over features describing a patient, to learn meaningful prototypical parts from structured data. These parts are represented as binary or discretized feature subsets. This allows the model to express prototypes in human-readable terms, enabling alignment with clinical language and case-based reasoning. Our proposed neural network is inherently interpretable and offers interpretable concept-based predictions by comparing the patient's description to learned prototypes in the latent space of the network. In experiments, we demonstrate that the model achieves classification performance competitive to widely used baseline models on medical benchmark datasets, while also offering transparency, bridging the gap between predictive performance and interpretability in clinical decision support.

Fetal ultrasound AI could transform prenatal care in low-resource settings, yet current foundation models exceed 300M visual parameters, precluding deployment on point-of-care devices. Standard knowledge distillation fails under such extreme capacity gaps (~26x), as compact students waste capacity mimicking architectural artifacts of oversized teachers. We introduce Selective Repulsive Knowledge Distillation, which decomposes contrastive KD into diagonal and off-diagonal components: matched pair alignment is preserved while the off-diagonal weight decays into negative values, repelling the student from the teacher's inter-class confusions and forcing discovery of architecturally native features. Our 11.4M parameter student surpasses the 304M-parameter FetalCLIP teacher on zero-shot HC18 biometry validity (88.6% vs. 83.5%) and brain sub-plane F1 (0.784 vs. 0.702), while running at 1.6 ms on iPhone 16 Pro, enabling real-time assistive AI on handheld ultrasound devices. Our code, models, and app are publicly available at https://github.com/numanai/MobileFetalCLIP.

Reaching for grasping, and manipulating objects are essential motor functions in everyday life. Decoding human motor intentions is a central challenge for rehabilitation and assistive technologies. This study focuses on predicting intentions by inferring movement direction and target location from multichannel electromyography (EMG) signals, and investigating how spatially and temporally accurate such information can be detected relative to movement onset. We present a computational pipeline that combines data-driven temporal segmentation with classical and deep learning classifiers in order to analyse EMG data recorded during the planning, early execution, and target contact phases of a delayed reaching task. Early intention prediction enables devices to anticipate user actions, improving responsiveness and supporting active motor recovery in adaptive rehabilitation systems. Random Forest achieves $80\%$ accuracy and Convolutional Neural Network $75\%$ accuracy across $25$ spatial targets, each separated by $14^\circ$ azimuth/altitude. Furthermore, a systematic evaluation of EMG channels, feature sets, and temporal windows demonstrates that motor intention can be efficiently decoded even with drastically reduced data. This work sheds light on the temporal and spatial evolution of motor intention, paving the way for anticipatory control in adaptive rehabilitation systems and driving advancements in computational approaches to motor neuroscience.

Introspection is a foundational cognitive ability, but its mechanism is not well understood. Recent work has shown that AI models can introspect. We study their mechanism of introspection, first extensively replicating Lindsey et al. (2025)'s thought injection detection paradigm in large open-source models. We show that these models detect injected representations via two separable mechanisms: (i) probability-matching (inferring from perceived anomaly of the prompt) and (ii) direct access to internal states. The direct access mechanism is content-agnostic: models detect that an anomaly occurred but cannot reliably identify its semantic content. The two model classes we study confabulate injected concepts that are high-frequency and concrete (e.g., "apple'"); for them correct concept guesses typically require significantly more tokens. This content-agnostic introspective mechanism is consistent with leading theories in philosophy and psychology.

Word Sense Disambiguation (WSD) remains a key challenge in Natural Language Processing (NLP), especially when dealing with rare or domain-specific senses that are often misinterpreted. While modern high-parameter Large Language Models (LLMs) such as GPT-4-Turbo have shown state-of-the-art WSD performance, their computational and energy demands limit scalability. This study investigates whether low-parameter LLMs (<4B parameters) can achieve comparable results through fine-tuning strategies that emphasize reasoning-driven sense identification. Using the FEWS dataset augmented with semi-automated, rationale-rich annotations, we fine-tune eight small-scale open-source LLMs (e.g. Gemma and Qwen). Our results reveal that Chain-of-Thought (CoT)-based reasoning combined with neighbour-word analysis achieves performance comparable to GPT-4-Turbo in zero-shot settings. Importantly, Gemma-3-4B and Qwen-3-4B models consistently outperform all medium-parameter baselines and state-of-the-art models on FEWS, with robust generalization to unseen senses. Furthermore, evaluation on the unseen "Fool Me If You Can'' dataset confirms strong cross-domain adaptability without task-specific fine-tuning. This work demonstrates that with carefully crafted reasoning-centric fine-tuning, low-parameter LLMs can deliver accurate WSD while substantially reducing computational and energy demands.

We present the Judge Reliability Harness, an open source library for constructing validation suites that test the reliability of LLM judges. As LLM based scoring is widely deployed in AI benchmarks, more tooling is needed to efficiently assess the reliability of these methods. Given a benchmark dataset and an LLM judge configuration, the harness generates reliability tests that evaluate both binary judgment accuracy and ordinal grading performance for free-response and agentic task formats. We evaluate four state-of-the-art judges across four benchmarks spanning safety, persuasion, misuse, and agentic behavior, and find meaningful variation in performance across models and perturbation types, highlighting opportunities to improve the robustness of LLM judges. No judge that we evaluated is uniformly reliable across benchmarks using our harness. For example, our preliminary experiments on judges revealed consistency issues as measured by accuracy in judging another LLM's ability to complete a task due to simple text formatting changes, paraphrasing, changes in verbosity, and flipping the ground truth label in LLM-produced responses. The code for this tool is available at: https://github.com/RANDCorporation/judge-reliability-harness

The emergence of generative AI models has dramatically expanded the availability and use of synthetic data across scientific, industrial, and policy domains. While these developments open new possibilities for data analysis, they also raise fundamental statistical questions about when synthetic data can be used in a valid, reliable, and principled manner. This paper reviews the current landscape of synthetic data generation and use from a statistical perspective, with the goal of clarifying the assumptions under which synthetic data can meaningfully support downstream discovery, inference, and prediction. We survey major classes of modern generative models, their intended use cases, and the benefits they offer, while also highlighting their limitations and characteristic failure modes. We additionally examine common pitfalls that arise when synthetic data are treated as surrogates for real observations, including biases from model misspecification, attenuated uncertainty, and difficulties in generalization. Building on these insights, we discuss emerging frameworks for the principled use of synthetic data. We conclude with practical recommendations, open problems, and cautions intended to guide both method developers and applied researchers.

Sheaf Neural Networks (SNNs) were introduced as an extension of Graph Convolutional Networks to address oversmoothing on heterophilous graphs by attaching a sheaf to the input graph and replacing the adjacency-based operator with a sheaf Laplacian defined by (learnable) restriction maps. Prior work motivates this design through theoretical properties of sheaf diffusion and the kernel of the sheaf Laplacian, suggesting that suitable non-identity restriction maps can avoid representations converging to constants across connected components. Since oversmoothing can also be mitigated through residual connections and normalization, we revisit a trivial sheaf construction to ask whether the additional complexity of learning restriction maps is necessary. We introduce an Identity Sheaf Network baseline, where all restriction maps are fixed to the identity, and use it to ablate the empirical improvements reported by sheaf-learning architectures. Across five popular heterophilic benchmarks, the identity baseline achieves comparable performance to a range of SNN variants. Finally, we introduce the Rayleigh quotient as a normalized measure for comparing oversmoothing across models and show that, in trained networks, the behavior predicted by the diffusion-based analysis of SNNs is not reflected empirically. In particular, Identity Sheaf Networks do not appear to suffer more significant oversmoothing than their SNN counterparts.

AI and Law research has encountered legal interpretation in different ways, in the context of its evolving approaches and methodologies. Research on expert system has focused on legal knowledge engineering, with the goal of ensuring that human-generated interpretations can be precisely transferred into knowledge-bases, to be consistently applied. Research on argumentation has aimed at representing the structure of interpretive arguments, as well as their dialectical interactions, to assess of the acceptability of interpretive claims within argumentation frameworks. Research on machine learning has focused on the automated generation of interpretive suggestions and arguments, through general and specialised language models, now being increasingly deployed in legal practice.

Estimating node similarity is a fundamental task in network analysis and graph-based machine learning, with applications in clustering, community detection, classification, and recommendation. We propose TopKGraphs, a method based on start-node-anchored random walks that bias transitions toward nodes with structurally similar neighborhoods, measured via Jaccard similarity. Rather than computing stationary distributions, walks are treated as stochastic neighborhood samplers, producing partial node rankings that are aggregated using robust rank aggregation to construct interpretable node-to-node affinity matrices. TopKGraphs provides a non-parametric, interpretable, and general-purpose representation of node similarity that can be applied in both network analysis and machine learning workflows. We evaluate the method on synthetic graphs (stochastic block models, Lancichinetti-Fortunato-Radicchi benchmark graphs), k-nearest-neighbor graphs from tabular datasets, and a curated high-confidence protein-protein interaction network. Across all scenarios, TopKGraphs achieves competitive or superior performance compared to standard similarity measures (Jaccard, Dice), a diffusion-based method (personalized PageRank), and an embedding-based approach (Node2Vec), demonstrating robustness in sparse, noisy, or heterogeneous networks. These results suggest that TopKGraphs is a versatile and interpretable tool for bridging simple local similarity measures with more complex embedding-based approaches, facilitating both data mining and network analysis applications.

This paper addresses the problem of Human Activity Recognition (HAR) using data from wearable inertial sensors. An important challenge in HAR is the model's generalization capabilities to new unseen individuals due to inter-subject variability, i.e., the same activity is performed differently by different individuals. To address this problem, we propose a novel deep adversarial framework that integrates the concept of inter-subject variability in the adversarial task, thereby encouraging subject-invariant feature representations and enhancing the classification performance in the HAR problem. Our approach outperforms previous methods in three well-established HAR datasets using a leave-one-subject-out (LOSO) cross-validation. Further results indicate that our proposed adversarial task effectively reduces inter-subject variability among different users in the feature space, and it outperforms adversarial tasks from previous works when integrated into our framework. Code: https://github.com/FranciscoCalatrava/EmbeddedSubjectVariability.git

Causal structure learning from observational data is central to many scientific and policy domains, but the time series setting common to many disciplines poses several challenges due to temporal dependence. In this paper we focus on score-based causal discovery for multivariate time series and introduce TS-BOSS, a time series extension of the recently proposed Best Order Score Search (BOSS) (Andrews et al. 2023). TS-BOSS performs a permutation-based search over dynamic Bayesian network structures while leveraging grow-shrink trees to cache intermediate score computations, preserving the scalability and strong empirical performance of BOSS in the static setting. We provide theoretical guarantees establishing the soundness of TS-BOSS under suitable assumptions, and we present an intermediate result that extends classical subgraph minimality results for permutation-based methods to the dynamic (time series) setting. Our experiments on synthetic data show that TS-BOSS is especially effective in high auto-correlation regimes, where it consistently achieves higher adjacency recall at comparable precision than standard constraint-based methods. Overall, TS-BOSS offers a high-performing, scalable approach for time series causal discovery and our results provide a principled bridge for extending sparsity-based, permutation-driven causal learning theory to dynamic settings.

Transformer architectures serve as the backbone for most modern Large Language Models, therefore their pretraining stability and convergence speed are of central concern. Motivated by the logical dependency of sequentially stacked layers, we propose Progressive Residual Warmup (ProRes) for language model pretraining. ProRes implements an "early layer learns first" philosophy by multiplying each layer's residual with a scalar that gradually warms up from 0 to 1, with deeper layers taking longer warmup steps. In this way, deeper layers wait for early layers to settle into a more stable regime before contributing to learning. We demonstrate the effectiveness of ProRes through pretraining experiments across various model scales, as well as normalization and initialization schemes. Comprehensive analysis shows that ProRes not only stabilizes pretraining but also introduces a unique optimization trajectory, leading to faster convergence, stronger generalization and better downstream performance. Our code is available at https://github.com/dandingsky/ProRes.

Writing efficient distributed code remains a labor-intensive and complex endeavor. To simplify application development, the Flexible Computational Science Infrastructure (FleCSI) framework offers a user-oriented, high-level programming interface that is built upon a task-based runtime model. Internally, FleCSI integrates state-of-the-art parallelization backends, including MPI and the asynchronous many-task runtimes (AMTRs) Legion and HPX, enabling applications to fully leverage asynchronous parallelism. In this work, we benchmark two applications using FleCSI's three backends on up to 1024 nodes, intending to quantify the advantages and overheads introduced by the AMTR backends. As representative applications, we select a simple Poisson solver and the multidimensional radiation hydrodynamics code HARD. In the communication-focused Poisson solver benchmark, FleCSI achieves over 97% parallel efficiency using the MPI backend under weak scaling on up to 131072 cores, indicating that only minimal overhead is introduced by its abstraction layer. While the Legion backend exhibits notable overheads and scaling limitations, the HPX backend introduces only marginal overhead compared to MPI+Kokkos. However, the scalability of the HPX backend is currently limited due to the usage of non-optimized HPX collective operations. In the computation-focused radiation hydrodynamics benchmarks, the performance gap between the MPI and HPX backends fades. On fewer than 64 nodes, the HPX backend outperforms MPI+Kokkos, achieving an average speedup of 1.31 under weak scaling and up to 1.27 under strong scaling. For the hydrodynamics-only HARD benchmark, the HPX backend demonstrates superior performance on fewer than 32 nodes, achieving speedups of up to 1.20 relative to MPI and up to 1.64 relative to MPI+Kokkos.

9-1-1 call-taking training requires mastery of over a thousand interdependent skills, covering diverse incident types and protocol-specific nuances. A nationwide labor shortage is already straining training capacity, but effective instruction still demands that trainers tailor objectives to each trainee's evolving competencies. This personalization burden is one that current practice cannot scale. Partnering with Metro Nashville Department of Emergency Communications (MNDEC), we propose PACE (Personalized Adaptive Curriculum Engine), a co-pilot system that augments trainer decision-making by (1) maintaining probabilistic beliefs over trainee skill states, (2) modeling individual learning and forgetting dynamics, and (3) recommending training scenarios that balance acquisition of new competencies with retention of existing ones. PACE propagates evidence over a structured skill graph to accelerate diagnostic coverage and applies contextual bandits to select scenarios that target gaps the trainee is prepared to address. Empirical results show that PACE achieves 19.50% faster time-to-competence and 10.95% higher terminal mastery compared to state-of-the-art frameworks. Co-pilot studies with practicing training officers further demonstrate a 95.45% alignment rate between PACE's and experts' pedagogical judgments on real-world cases. Under estimation, PACE cuts turnaround time to merely 34 seconds from 11.58 minutes, up to 95.08% reduction.

Test-time adaptation offers a promising avenue for improving reasoning performance in large language models without additional supervision, but existing approaches often apply a uniform optimization objective across all inputs, leading to inefficient or unstable adaptation on heterogeneous reasoning problems. We propose DiSCTT, a difficulty-aware, consensus-guided self-curriculum framework that dynamically allocates test-time optimization strategies based on instance-level epistemic uncertainty estimated from agreement among sampled reasoning trajectories. Inputs with high consensus are consolidated via supervised fine-tuning using majority-agreed solutions as pseudo-labels, while low-consensus inputs are optimized via reinforcement learning with a consensus-regularized objective that encourages diversity under relevance constraints. Across a broad suite of mathematical and general reasoning benchmarks, DiSCTT consistently outperforms strong test-time adaptation baselines, achieving higher accuracy with reduced variance and substantially lower computation and wall-clock training times. These results demonstrate that explicitly accounting for instance difficulty and uncertainty enables more stable, efficient, and effective test-time adaptation for reasoning models.

Model merging is a scalable alternative to multi-task training that combines the capabilities of multiple specialised models into a single model. This is particularly attractive for large speech foundation models, which are typically adapted through domain-specific fine-tuning, resulting in multiple customised checkpoints, for which repeating full fine-tuning when new data becomes available is computationally prohibitive. In this work, we study model merging for multi-domain ASR and benchmark 11 merging algorithms for 10 European Portuguese domains, evaluating in-domain accuracy, robustness under distribution shift, as well as English and multilingual performance. We further propose BoostedTSV-M, a new merging algorithm based on TSV-M that mitigates rank collapse via singular-value boosting and improves numerical stability. Overall, our approach outperforms full fine-tuning on European Portuguese while preserving out-of-distribution generalisation in a single model.

Retrieval-augmented generation (RAG) for long-context question answering is bottlenecked by inference-time prefilling over large retrieved contexts. A common strategy is to precompute key-value (KV) caches for individual documents and selectively recompute a small subset of tokens to restore global causal dependencies, but existing methods rely on heuristics or representation discrepancies without modeling whether selected tokens can effectively influence generation. We cast selective KV recomputation as an information flow problem and show that a simple attention-norm signal from the query reliably identifies tokens that are both semantically relevant and structurally positioned to propagate information, when computed under an inference-consistent RoPE geometry. We therefore reconstruct global positional assignments for retrieved chunks and introduce an information-flow-guided chunk reordering strategy. Experiments on LLM and VLM benchmarks demonstrate consistent gains over prior methods under comparable efficiency budgets.

Chess engines passed human strength years ago, but they still don't play like humans. A grandmaster under clock pressure blunders in ways a club player on a hot streak never would. Conventional engines capture none of this. This paper proposes a personality x psyche decomposition to produce behavioral variability in chess play, drawing on patterns observed in human games. Personality is static -- a preset that pins down the engine's character. Psyche is dynamic -- a bounded scalar ψ_t \in [-100, +100], recomputed from five positional factors after every move. These two components feed into an audio-inspired signal chain (noise gate, compressor/expander, five-band equalizer, saturation limiter) that reshapes move probability distributions on the fly. The chain doesn't care what engine sits behind it: any system that outputs move probabilities will do. It needs no search and carries no state beyond ψ_t. I test the framework across 12,414 games against Maia2-1100, feeding it two probability sources that differ by ~2,800x in training data. Both show the same monotonic gradient in top-move agreement (~20-25 pp spread from stress to overconfidence), which tells us the behavioral variation comes from the signal chain, not from the model underneath. When the psyche runs overconfident, the chain mostly gets out of the way (66% agreement with vanilla Maia2). Under stress, the competitive score falls from 50.8% to 30.1%. The patterns are reminiscent of tilt and overconfidence as described in human play, but I should be upfront: this study includes no human-subject validation.

Being able to understand information is a key factor for a self-determined life and society. It is also very important for participating in democratic processes. The study of automatic text simplification is often limited by the availability of high quality material for the training and evaluation on automatic simplifiers. This is true for English, but more so for less resourced languages like Spanish, Catalan and Italian. In order to fill this gap, we present a corpus of original texts for these 3 languages, with high quality simplification produced by human experts in text simplification. It was developed within the iDEM project to assess the impact of Easy-to-Read (E2R) language for democratic participation. The original texts were compiled from domains related to this topic. The corpus includes different text types, selected based on relevance, copyright availability, and ethical standards. All texts were simplified to E2R level. The corpus is particularity valuable because it includes the first annotated corpus of its kind for the Catalan language. It also represents a noteworthy contribution for Spanish and Italian, offering high-quality, human-annotated language resources that are rarely available in these domains. The corpus will be made freely accessible to the public.

The landscape of AI coding assistance is undergoing a fundamental shift from complex IDE plugins to versatile, terminal-native agents. Operating directly where developers manage source control, execute builds, and deploy environments, CLI-based agents offer unprecedented autonomy for long-horizon development tasks. In this paper, we present OPENDEV, an open-source, command-line coding agent engineered specifically for this new paradigm. Effective autonomous assistance requires strict safety controls and highly efficient context management to prevent context bloat and reasoning degradation. OPENDEV overcomes these challenges through a compound AI system architecture with workload-specialized model routing, a dual-agent architecture separating planning from execution, lazy tool discovery, and adaptive context compaction that progressively reduces older observations. Furthermore, it employs an automated memory system to accumulate project-specific knowledge across sessions and counteracts instruction fade-out through event-driven system reminders. By enforcing explicit reasoning phases and prioritizing context efficiency, OPENDEV provides a secure, extensible foundation for terminal-first AI assistance, offering a blueprint for robust autonomous software engineering.

Equivariant Graph Neural Networks (GNNs) are essential for physically consistent molecular simulations but suffer from high computational costs and memory bottlenecks, especially with high-order representations. While low-bit quantization offers a solution, applying it naively to rotation-sensitive features destroys the SO(3)-equivariant structure, leading to significant errors and violations of conservation laws. To address this issue, in this work, we propose a Geometric-Aware Quantization (GAQ) framework that compresses and accelerates equivariant models while rigorously preserving continuous symmetry in discrete spaces. Our approach introduces three key contributions: (1) a Magnitude-Direction Decoupled Quantization (MDDQ) scheme that separates invariant lengths from equivariant orientations to maintain geometric fidelity; (2) a symmetry-aware training strategy that treats scalar and vector features with distinct quantization schedules; and (3) a robust attention normalization mechanism to stabilize gradients in low-bit regimes. Experiments on the rMD17 benchmark demonstrate that our W4A8 models match the accuracy of FP32 baselines (9.31 meV vs. 23.20 meV) while reducing Local Equivariance Error (LEE) by over 30x compared to naive quantization. On consumer hardware, GAQ achieves 2.39x inference speedup and 4x memory reduction, enabling stable, energy-conserving molecular dynamics simulations for nanosecond timescales.

While globally optimal empirical risk minimization (ERM) decision trees have become computationally feasible and empirically successful, rigorous theoretical guarantees for their statistical performance remain limited. In this work, we develop a comprehensive statistical theory for ERM trees under random design in both high-dimensional regression and classification. We first establish sharp oracle inequalities that bound the excess risk of the ERM estimator relative to the best possible approximation achievable by any tree with at most $L$ leaves, thereby characterizing the interpretability-accuracy trade-off. We derive these results using a novel uniform concentration framework based on empirically localized Rademacher complexity. Furthermore, we derive minimax optimal rates over a novel function class: the piecewise sparse heterogeneous anisotropic Besov (PSHAB) space. This space explicitly captures three key structural features encountered in practice: sparsity, anisotropic smoothness, and spatial heterogeneity. While our main results are established under sub-Gaussianity, we also provide robust guarantees that hold under heavy-tailed noise settings. Together, these findings provide a principled foundation for the optimality of ERM trees and introduce empirical process tools broadly applicable to other highly adaptive, data-driven procedures.

Many models of opinion dynamics include measures of distance between opinions. Such models are susceptible to boundary effects where the choice of the topology of the opinion space may influence the dynamics. In this paper we study an opinion dynamics model following the seminal model by Axelrod, with the goal of understanding the effect of a toroidal opinion space. To do this we systematically compare two versions of the model: one with toroidal opinion space and one with cubic opinion space. In their most basic form the two versions of our model result in similar dynamics (consensus is attained eventually). However, as we include bounded confidence and eventually per agent weighting of opinion elements the dynamics become quite contrasting. The toroidal opinion space consistently allows for a greater number of groups in steady state than the cubic opinion space model. Furthermore, the outcome of the dynamics in the toroidal opinion space model are more sensitive to the inclusion of extensions than in the cubic opinion space model.

Four distinct admissibility geometries govern sequential and distribution-free inference: Blackwell risk dominance over convex risk sets, anytime-valid admissibility within the nonnegative supermartingale cone, marginal coverage validity over exchangeable prediction sets, and Cesàro approachability (CAA) admissibility, which reaches the risk-set boundary via approachability-style arguments rather than explicit priors. We prove a criterion separation theorem: the four classes of admissible procedures are pairwise non-nested. Each geometry carries a different certificate of optimality: a supporting-hyperplane prior (Blackwell), a nonnegative supermartingale (anytime-valid), an exchangeability rank (coverage), or a Cesàro steering argument (CAA). Martingale coherence is necessary for Blackwell admissibility and necessary and sufficient for anytime-valid admissibility within e-processes, but is not sufficient for Blackwell admissibility and is not necessary for coverage validity or CAA-admissibility. All four criteria share a common optimization template (minimize Bayesian risk subject to a feasibility constraint), but the constraint sets operate over different spaces, partial orders, and performance metrics, making them geometrically incompatible. Admissibility is irreducibly criterion-relative.

Financial datasets often suffer from bias that can lead to unfair decision-making in automated systems. In this work, we propose FairFinGAN, a WGAN-based framework designed to generate synthetic financial data while mitigating bias with respect to the protected attribute. Our approach incorporates fairness constraints directly into the training process through a classifier, ensuring that the synthetic data is both fair and preserves utility for downstream predictive tasks. We evaluate our proposed model on five real-world financial datasets and compare it with existing GAN-based data generation methods. Experimental results show that our approach achieves superior fairness metrics without significant loss in data utility, demonstrating its potential as a tool for bias-aware data generation in financial applications.

Time-series clustering is a fundamental tool for pattern discovery, yet existing explainability methods, primarily based on feature attribution or metadata, fail to identify the transitions that move an instance across cluster boundaries. While Counterfactual Explanations (CEs) identify the minimal temporal perturbations required to alter the prediction of a model, they have been mostly confined to supervised settings. This paper introduces GALACTIC, the first unified framework to bridge local and global counterfactual explainability for unsupervised time-series clustering. At instance level (local), GALACTIC generates perturbations via a cluster-aware optimization objective that respects the target and underlying cluster assignments. At cluster level (global), to mitigate cognitive load and enhance interpretability, we formulate a representative CE selection problem. We propose a Minimum Description Length (MDL) objective to extract a non-redundant summary of global explanations that characterize the transitions between clusters. We prove that our MDL objective is supermodular, which allows the corresponding MDL reduction to be framed as a monotone submodular set function. This enables an efficient greedy selection algorithm with provable $(1-1/e)$ approximation guarantees. Extensive experimental evaluation on the UCR Archive demonstrates that GALACTIC produces significantly sparser local CEs and more concise global summaries than state-of-the-art baselines adapted for our problem, offering the first unified approach for interpreting clustered time-series through counterfactuals.

In clinical prediction settings the importance of a high-dimensional feature like genomics is often assessed by evaluating the change in predictive performance when adding it to a set of traditional clinical variables. This approach is questionable, because it does not account for collinearity nor known directionality of dependencies between variables. We suggest to use asymmetric Shapley values as a more suitable alternative to quantify feature importance in the context of a mixed-dimensional prediction model. We focus on a setting that is particularly relevant in clinical prediction: disease state as a mediating variable for genomic effects, with additional confounders for which the direction of effects may be unknown. We derive efficient algorithms to compute local and global asymmetric Shapley values for this setting. The former are shown to be very useful for inference, whereas the latter provide interpretation by decomposing any predictive performance metric into contributions of the features. Throughout, we illustrate our framework by a leading example: the prediction of progression-free survival for colorectal cancer patients.

Punctuation restoration is essential for improving the readability and downstream utility of automatic speech recognition (ASR) outputs, yet remains underexplored for Persian despite its importance. We introduce PersianPunc, a large-scale, high-quality dataset of 17 million samples for Persian punctuation restoration, constructed through systematic aggregation and filtering of existing textual resources. We formulate punctuation restoration as a token-level sequence labeling task and fine-tune ParsBERT to achieve strong performance. Through comparative evaluation, we demonstrate that while large language models can perform punctuation restoration, they suffer from critical limitations: over-correction tendencies that introduce undesired edits beyond punctuation insertion (particularly problematic for speech-to-text pipelines) and substantially higher computational requirements. Our lightweight BERT-based approach achieves a macro-averaged F1 score of 91.33% on our test set while maintaining efficiency suitable for real-time applications. We make our dataset (https://huggingface.co/datasets/MohammadJRanjbar/persian-punctuation-restoration) and model (https://huggingface.co/MohammadJRanjbar/parsbert-persian-punctuation) publicly available to facilitate future research in Persian NLP and provide a scalable framework applicable to other morphologically rich, low-resource languages.

While existing audio watermarking techniques have achieved strong robustness against traditional digital signal processing (DSP) attacks, they remain vulnerable to neural resynthesis. This occurs because modern neural audio codecs act as semantic filters and discard the imperceptible waveform variations used in prior watermarking methods. To address this limitation, we propose Latent-Mark, the first zero-bit audio watermarking framework designed to survive semantic compression. Our key insight is that robustness to the encode-decode process requires embedding the watermark within the codec's invariant latent space. We achieve this by optimizing the audio waveform to induce a detectable directional shift in its encoded latent representation, while constraining perturbations to align with the natural audio manifold to ensure imperceptibility. To prevent overfitting to a single codec's quantization rules, we introduce Cross-Codec Optimization, jointly optimizing the waveform across multiple surrogate codecs to target shared latent invariants. Extensive evaluations demonstrate robust zero-shot transferability to unseen neural codecs, achieving state-of-the-art resilience against traditional DSP attacks while preserving perceptual imperceptibility. Our work inspires future research into universal watermarking frameworks capable of maintaining integrity across increasingly complex and diverse generative distortions.

Assessing whether an article supports an assertion is essential for hallucination detection and claim verification. While large language models (LLMs) have the potential to automate this task, achieving strong performance requires frontier models such as GPT-5 that are prohibitively expensive to deploy at scale. To efficiently perform biomedical evidence attribution, we present Med-V1, a family of small language models with only three billion parameters. Trained on high-quality synthetic data newly developed in this study, Med-V1 substantially outperforms (+27.0% to +71.3%) its base models on five biomedical benchmarks unified into a verification format. Despite its smaller size, Med-V1 performs comparably to frontier LLMs such as GPT-5, along with high-quality explanations for its predictions. We use Med-V1 to conduct a first-of-its-kind use case study that quantifies hallucinations in LLM-generated answers under different citation instructions. Results show that the format instruction strongly affects citation validity and hallucination, with GPT-5 generating more claims but exhibiting hallucination rates similar to GPT-4o. Additionally, we present a second use case showing that Med-V1 can automatically identify high-stakes evidence misattributions in clinical practice guidelines, revealing potentially negative public health impacts that are otherwise challenging to identify at scale. Overall, Med-V1 provides an efficient and accurate lightweight alternative to frontier LLMs for practical and real-world applications in biomedical evidence attribution and verification tasks. Med-V1 is available at https://github.com/ncbi-nlp/Med-V1.

Spatio-temporal kriging aims to infer signals at unobserved locations from observed sensors and is critical to applications such as transportation and environmental monitoring. In practice, however, observed sensors themselves often exhibit heterogeneous missingness, forcing inductive kriging models to rely on crudely imputed inputs. This setting brings three key challenges: (1) it is unclear whether an value is a true signal or a missingness-induced artifact; (2) missingness is highly heterogeneous across sensors and time; (3) missing observations distort the local spatio-temporal structure. To address these issues, we propose Uniform Inductive Spatio-Temporal Kriging (UniSTOK), a plug-and-play framework that enhances existing inductive kriging backbones under missing observation. Our framework forms a dual-branch input consisting of the original observations and a jigsaw-augmented counterpart that synthesizes proxy signals only at missing entries. The two branches are then processed in parallel by a shared spatio-temporal backbone with explicit missingness mask modulation. Their outputs are finally adaptively fused via dual-channel attention. Experiments on multiple real-world datasets under diverse missing patterns demonstrate consistent and significant improvements.

Large language models show that simple autoregressive training can yield scalable and coherent generation, but extending this paradigm to speech remains challenging due to the entanglement of semantic and acoustic information. Most existing speech language models rely on text supervision, hierarchical token streams, or complex hybrid architectures, departing from the single-stream generative pretraining paradigm that has proven effective in text. In this work, we introduce WavSLM, a speech language model trained by quantizing and distilling self-supervised WavLM representations into a single codebook and optimizing an autoregressive next-chunk prediction objective. WavSLM jointly models semantic and acoustic information within a single token stream without text supervision or text pretraining. Despite its simplicity, it achieves competitive performance on consistency benchmarks and speech generation while using fewer parameters, less training data, and supporting streaming inference. Demo samples are available at https://lucadellalib.github.io/wavslm-web/.

Offline reinforcement learning (RL) allows robots to learn from offline datasets without risky exploration. Yet, offline RL's performance often hinges on a brittle trade-off between (1) return maximization, which can push policies outside the dataset support, and (2) behavioral constraints, which typically require sensitive hyperparameter tuning. Latent steering offers a structural way to stay within the dataset support during RL, but existing offline adaptations commonly approximate action values using latent-space critics learned via indirect distillation, which can lose information and hinder convergence. We propose Latent Policy Steering (LPS), which enables high-fidelity latent policy improvement by backpropagating original-action-space Q-gradients through a differentiable one-step MeanFlow policy to update a latent-action-space actor. By eliminating proxy latent critics, LPS allows an original-action-space critic to guide end-to-end latent-space optimization, while the one-step MeanFlow policy serves as a behavior-constrained generative prior. This decoupling yields a robust method that works out-of-the-box with minimal tuning. Across OGBench and real-world robotic tasks, LPS achieves state-of-the-art performance and consistently outperforms behavioral cloning and strong latent steering baselines.

We introduce WebChain, the largest open-source dataset of human-annotated trajectories on real-world websites, designed to accelerate reproducible research in web agents. It contains 31,725 trajectories and 318k steps, featuring a core Triple Alignment of visual, structural, and action data to provide rich, multi-modal supervision. The data is collected via a scalable pipeline that ensures coverage of complex, high-value tasks often missed by synthetic methods. Leveraging this dataset, we propose a Dual Mid-Training recipe that decouples spatial grounding from planning, achieving state-of-the-art performance on our proposed WebChainBench and other public GUI benchmarks. Our work provides the data and insights necessary to build and rigorously evaluate the next generation of scalable web agents.

Recent advances in large language models (LLMs) have enabled agentic systems for sequential decision-making. Such agents must perceive their environment, reason across multiple time steps, and take actions that optimize long-term objectives. However, existing web agents struggle on complex, long-horizon tasks due to limited in-context memory for tracking history, weak planning abilities, and greedy behaviors that lead to premature termination. To address these challenges, we propose STRUCTUREDAGENT, a hierarchical planning framework with two core components: (1) an online hierarchical planner that uses dynamic AND/OR trees for efficient search and (2) a structured memory module that tracks and maintains candidate solutions to improve constraint satisfaction in information-seeking tasks. The framework also produces interpretable hierarchical plans, enabling easier debugging and facilitating human intervention when needed. Our results on WebVoyager, WebArena, and custom shopping benchmarks show that STRUCTUREDAGENT improves performance on long-horizon web-browsing tasks compared to standard LLM-based agents.

AI safety via debate and reinforcement learning from AI feedback (RLAIF) are both proposed methods for scalable oversight of advanced AI systems, yet no formal framework relates them or characterizes when debate offers an advantage. We analyze this by parameterizing debate's value through the geometry of knowledge divergence between debating models. Using principal angles between models' representation subspaces, we prove that the debate advantage admits an exact closed form. When models share identical training corpora, debate reduces to RLAIF-like where a single-agent method recovers the same optimum. When models possess divergent knowledge, debate advantage scales with a phase transition from quadratic regime (debate offers negligible benefit) to linear regime (debate is essential). We classify three regimes of knowledge divergence (shared, one-sided, and compositional) and provide existence results showing that debate can achieve outcomes inaccessible to either model alone, alongside a negative result showing that sufficiently strong adversarial incentives cause coordination failure in the compositional regime, with a sharp threshold separating effective from ineffective debate. We offer the first formal connection between debate and RLAIF, a geometric foundation for understanding when adversarial oversight protocols are justified, and connection to the problem of eliciting latent knowledge across models with complementary information.

Large language models (LLMs) achieve promising performance, yet their ability to reason remains poorly understood. Existing evaluations largely emphasize task-level accuracy, often conflating pattern matching with reasoning capability. We present X-RAY, an explainable reasoning analysis system that maps the LLM reasoning capability using calibrated, formally verified probes. We model reasoning capability as a function of extractable \textit{structure}, operationalized through formal properties such as constraint interaction, reasoning depth, and solution-space geometry. X-Ray generates probes via formal tools with controlled structural variations, enabling precise isolation of incremental structural information through formal calibration and verification. We evaluate state-of-the-art LLMs on problems ranging from junior-level to advanced in mathematics, physics, and chemistry. Our analysis reveals a systematic asymmetry in LLM reasoning: models are relatively robust to constraint refinement, where additional conditions shrink an existing solution space, but degrade sharply under solution-space restructuring, where modifications alter the underlying structural form of the solution manifold. Moreover, calibrated formal probes differentiate models that appear indistinguishable on standard benchmarks and reveal failure modes that are structurally interpretable rather than opaque. Beyond evaluation, our framework is contamination-free and supports the training and testing of reasoning models.

Finding patient subgroups with similar characteristics is crucial for personalized decision-making in various disciplines such as healthcare and policy evaluation. While most existing approaches rely on unsupervised clustering methods, there is a growing trend toward using supervised clustering methods that identify operationalizable subgroups in the context of a specific outcome of interest. We propose Bayesian Supervised Causal Clustering (BSCC), with treatment effect as outcome to guide the clustering process. BSCC identifies homogenous subgroups of individuals who are similar in their covariate profiles as well as their treatment effects. We evaluate BSCC on simulated datasets as well as real-world dataset from the third International Stroke Trial to assess the practical usefulness of the framework.

Recent studies have observed that intermediate layers of foundation models often yield more discriminative representations than the final layer. While initially attributed to autoregressive pretraining, this phenomenon has also been identified in models trained via supervised and discriminative self-supervised objectives. In this paper, we conduct a comprehensive study to analyze the behavior of intermediate layers in pretrained vision transformers. Through extensive linear probing experiments across a diverse set of image classification benchmarks, we find that distribution shift between pretraining and downstream data is the primary cause of performance degradation in deeper layers. Furthermore, we perform a fine-grained analysis at the module level. Our findings reveal that standard probing of transformer block outputs is suboptimal; instead, probing the activation within the feedforward network yields the best performance under significant distribution shift, whereas the normalized output of the multi-head self-attention module is optimal when the shift is weak.

Large language models (LLMs) can be used to support software development tasks, e.g., through code completion or code generation. However, their effectiveness drops significantly when considering less popular programming languages such as domain-specific languages (DSLs). In this paper, we propose a generic framework for evaluating the capabilities of LLMs generating DSL code from textual specifications. The generated code is assessed from the perspectives of well-formedness and correctness. This framework is applied to a particular type of DSL, constraint languages, focusing our experiments on OCL and Alloy and comparing their results to those achieved for Python, a popular general-purpose programming language. Experimental results show that, in general, LLMs have better performance for Python than for OCL and Alloy. LLMs with smaller context windows such as open-source LLMs may be unable to generate constraint-related code, as this requires managing both the constraint and the domain model where it is defined. Moreover, some improvements to the code generation process such as code repair (asking an LLM to fix incorrect code) or multiple attempts (generating several candidates for each coding task) can improve the quality of the generated code. Meanwhile, other decisions like the choice of a prompt template have less impact. All these dimensions can be systematically analyzed using our evaluation framework, making it possible to decide the most effective way to set up code generation for a particular type of task.

In the landscape of modern machine learning, frozen pre-trained models provide stability and efficiency but often underperform on specific tasks due to mismatched data distributions. This paper introduces the Whisperer, a novel visual prompting framework that learns diffusion-based preprocessors to adapt inputs in pixel space, effectively "whispering" enhancements to frozen downstream models like EasyOCR. By framing the process as behavioral cloning of stochastically discovered improvement policies, our method achieves an 8% absolute (10.6% relative) reduction in Character Error Rate (CER) on a challenging dataset of 300k degraded synthetic text images, surpassing hand-engineered baselines such as CLAHE. The key innovation is a four-stage training curriculum that uses behavioral cloning to amplify "lucky" improvements discovered through the stochastic exploration of a partially trained diffusion model. This approach is highly sample-efficient and avoids the pitfalls of traditional reinforcement learning. Crucially, we frame this not as naive reinforcement learning, but as behavioral cloning of an exploration policy: we stochastically sample intermediate diffusion outputs, select those that improve CER by chance, and then train the model to reproduce them. This bootstrapping curriculum (4 stages over 60 GPU-hours) amplifies random successes into a systematic strategy. In summary, by whispering to the frozen OCR through its inputs, we improve an imperfect classifier without touching its weights.

Multimodal sarcasm detection requires resolving pragmatic incongruity across textual, acoustic, and visual cues through cross-modal reasoning. To enable robust sarcasm reasoning with foundation models, we propose SarcasmMiner, a reinforcement learning based post-training framework that resists hallucination in multimodal reasoning. We reformulate sarcasm detection as structured reasoning and adopt a dual-track distillation strategy: high-quality teacher trajectories initialize the student model, while the full set of trajectories trains a generative reward model (GenRM) to evaluate reasoning quality. The student is optimized with group relative policy optimization (GRPO) using decoupled rewards for accuracy and reasoning quality. On MUStARD++, SarcasmMiner increases F1 from 59.83% (zero-shot), 68.23% (supervised finetuning) to 70.22%. These findings suggest that reasoning-aware reward modeling enhances both performance and multimodal grounding.

We present the Multilingual Cloud Corpus, the first national-scale, parallel, multimodal linguistic dataset of Bangladesh's ethnic and indigenous languages. Despite being home to approximately 40 minority languages spanning four language families, Bangladesh has lacked a systematic, cross-family digital corpus for these predominantly oral, computationally "zero resource" varieties, 14 of which are classified as endangered. Our corpus comprises 85792 structured textual entries, each containing a Bengali stimulus text, an English translation, and an IPA transcription, together with approximately 107 hours of transcribed audio recordings, covering 42 language varieties from the Tibeto-Burman, Indo-European, Austro-Asiatic, and Dravidian families, plus two genetically unclassified languages. The data were collected through systematic fieldwork over 90 days across nine districts of Bangladesh, involving 16 data collectors, 77 speakers, and 43 validators, following a predefined elicitation template of 2224 unique items organized at three levels of linguistic granularity: isolated lexical items (475 words across 22 semantic domains), grammatical constructions (887 sentences across 21 categories including verbal conjugation paradigms), and directed speech (862 prompts across 46 conversational scenarios). Post-field processing included IPA transcription by 10 linguists with independent adjudication by 6 reviewers. The complete dataset is publicly accessible through the Multilingual Cloud platform (multiling.cloud), providing searchable access to annotated audio and textual data for all documented varieties. We describe the corpus design, fieldwork methodology, dataset structure, and per-language coverage, and discuss implications for endangered language documentation, low-resource NLP, and digital preservation in linguistically diverse developing countries.

Recent studies have demonstrated that incorporating auxiliary information, such as speaker voiceprint or visual cues, can substantially improve Speech Enhancement (SE) performance. However, single-channel methods often yield suboptimal results in low signal-to-noise ratio (SNR) conditions, when there is high reverberation, or in complex scenarios involving dynamic speakers, overlapping speech, or non-stationary noise. To address these issues, we propose a novel Visual-Informed Neural Beamforming Network (VI-NBFNet), which integrates microphone array signal processing and deep neural networks (DNNs) using multimodal input features. The proposed network leverages a pretrained visual speech recognition model to extract lip movements as input features, which serve for voice activity detection (VAD) and target speaker identification. The system is intended to handle both static and moving speakers by introducing a supervised end-to-end beamforming framework equipped with an attention mechanism. The experimental results demonstrated that the proposed audiovisual system has achieved better SE performance and robustness for both stationary and dynamic speaker scenarios, compared to several baseline methods.

Automatic speech recognition (ASR) systems are predominantly evaluated using the Word Error Rate (WER). However, raw token-level metrics fail to capture semantic fidelity and routinely obscures the `diversity tax', the disproportionate burden on marginalized and atypical speaker due to systematic recognition failures. In this paper, we explore the limitations of relying solely on lexical counts by systematically evaluating a broader class of non-linear and semantic metrics. To enable rigorous model auditing, we introduce the sample difficulty index (SDI), a novel metric that quantifies how intrinsic demographic and acoustic factors drive model failure. By mapping SDI on data cartography, we demonstrate that metrics EmbER and SemDist expose hidden systemic biases and inter-model disagreements that WER ignores. Finally, our findings are the first steps towards a robust audit framework for prospective safety analysis, empowering developers to audit and mitigate ASR disparities prior to deployment.

Transformer-based large language models are increasingly constrained by data movement as communication bandwidth drops sharply beyond the chip boundary. Wafer-scale integration using wafer-on-wafer hybrid bonding alleviates this limitation by providing ultra-high bandwidth between reticles on bonded wafers. In this paper, we investigate how the physical placement of reticles on wafers influences the achievable network topology and the resulting communication performance. Starting from a 2D mesh-like baseline, we propose four reticle placements (Aligned, Interleaved, Rotated, and Contoured) that improve throughput by up to 250%, reduce latency by up to 36%, and decrease energy per transmitted byte by up to 38%.

Accurate short-term wind power forecasting is essential for grid dispatch and market operations, yet centralising turbine data raises privacy, cost, and heterogeneity concerns. We propose a two-stage federated learning framework that first clusters turbines by long-term behavioural statistics using Double Roulette Selection (DRS) initialisation with recursive Auto-split refinement, and then trains cluster-specific LSTM models via FedAvg. Experiments on 400 stand-alone turbines in Denmark show that DRS-auto discovers behaviourally coherent groups and achieves competitive forecasting accuracy while preserving data locality. Behaviour-aware grouping consistently outperforms geographic partitioning and matches strong k-means++ baselines, suggesting a practical privacy-friendly solution for heterogeneous distributed turbine fleets.

VietJobs is the first large-scale, publicly available corpus of Vietnamese job advertisements, comprising 48,092 postings and over 15 million words collected from all 34 provinces and municipalities across Vietnam. The dataset provides extensive linguistic and structured information, including job titles, categories, salaries, skills, and employment conditions, covering 16 occupational domains and multiple employment types (full-time, part-time, and internship). Designed to support research in natural language processing and labour market analytics, VietJobs captures substantial linguistic, regional, and socio-economic diversity. We benchmark several generative large language models (LLMs) on two core tasks: job category classification and salary estimation. Instruction-tuned models such as Qwen2.5-7B-Instruct and Llama-SEA-LION-v3-8B-IT demonstrate notable gains under few-shot and fine-tuned settings, while highlighting challenges in multilingual and Vietnamese-specific modelling for structured labour market prediction. VietJobs establishes a new benchmark for Vietnamese NLP and offers a valuable foundation for future research on recruitment language, socio-economic representation, and AI-driven labour market analysis. All code and resources are available at: https://github.com/VinNLP/VietJobs.

Empirical and LLM-based research in model-driven engineering increasingly relies on datasets of software models, for instance, to train or evaluate machine learning techniques for modeling support. These datasets have a significant impact on solution performance; hence, they should be treated and assessed as first-class artifacts. However, such datasets are typically collected or created ad hoc and without guarantees of their quality for the specific task for which they are used. This limits the comparability of results between studies, obscures dataset quality and representativeness, and leads to weak reproducibility and potential bias. In this work, we propose a benchmarking framework for model datasets (i.e., benchmarking the dataset itself). Benchmarking datasets involves systematically measuring their quality, representativeness, and suitability for specific tasks. To this end, we propose a Benchmark Platform for MDE that provides a unified infrastructure for systematically assessing and comparing datasets of software models across languages and formats, using defined criteria and metrics.

Applications requiring real-time processing of large volumes of data have been the main driver for rethinking the traditional cloud, giving rise to novel cloud models. Distributed cloud (DC) is a model that allows users to dynamically create and dispose of strategically located ad-hoc clouds that contain resources best tailored to their needs. It is essential for this model to provide a high degree of observability for it to be viable in real-world scenarios. In this paper, we present the design and implementation of a monitoring system that collects metrics from DCs and makes them accessible to diverse clients. Agents running on nodes are responsible for collecting machine-, container-, and application-level metrics. During the health-check protocol, that data is transferred from the node to the DC's control plane running inside the cloud. There, it is persisted and served via multiple APIs, including a streaming API. Moreover, node metrics are aggregated for every DC in order to provide a more comprehensive view of the system's state.

As a key form in online social platforms, group chat is a popular space for interest exchange or problem-solving, but its effectiveness is often hindered by inactivity and management challenges. While recent large language models (LLMs) have powered impressive one-to-one conversational agents, their seamlessly integration into multi-participant conversations remains unexplored. To address this gap, we introduce GCAgent, an LLM-driven system for enhancing group chats communication with both entertainment- and utility-oriented dialogue agents. The system comprises three tightly integrated modules: Agent Builder, which customizes agents to align with users' interests; Dialogue Manager, which coordinates dialogue states and manage agent invocations; and Interface Plugins, which reduce interaction barriers by three distinct tools. Through extensive experiment, GCAgent achieved an average score of 4.68 across various criteria and was preferred in 51.04\% of cases compared to its base model. Additionally, in real-world deployments over 350 days, it increased message volume by 28.80\%, significantly improving group activity and engagement. Overall, this work presents a practical blueprint for extending LLM-based dialogue agent from one-party chats to multi-party group scenarios.

Source-Free Cross-Domain Few-Shot Learning (SF-CDFSL) focuses on fine-tuning with limited training data from target domains (e.g., medical or satellite images), where CLIP has recently shown promising results due to its generalizability to downstream tasks. Current works indicate CLIP's text encoder is more suitable for cross-domain tasks, however, we find that \textbf{removing certain middle layers of the text encoder can effectively improve performance in SF-CDFSL}, which we call the Lost Layers. In this paper, we delve into this phenomenon for a deeper understanding. We discover that instead of being harmful for the SF-CDFSL task, the information in these layers is actually beneficial, but visual gaps prevent this useful information from being fully utilized, making these layers seem redundant. Based on this understanding, unlike current works that simply remove these layers, we propose a method to teachs the model to \textbf{re-utilize} information in these lost layers at both the layer and encoder levels, guiding the re-learning of the visual branch under domain shifts. Our approach effectively addresses the issue of underutilized information in the text encoder. Extensive experiments across various settings, backbones (CLIP, SigLip, PE-Core), and tasks (4 CDFSL datasets and 10 Meta-dataset datasets) demonstrate the effectiveness of our method. Code is available at https://github.com/zhenyuZ-HUST/CVPR26-VtT.

Neural reasoners such as Tiny Recursive Models (TRMs) solve complex problems by combining neural backbones with specialized inference schemes. Such inference schemes have been a central component of stochastic reasoning systems, where inference rules are applied to a stochastic model to derive answers to complex queries. In this work, we bridge these two paradigms by introducing Recursive Inference Machines (RIMs), a neural reasoning framework that explicitly incorporates recursive inference mechanisms inspired by classical inference engines. We show that TRMs can be expressed as an instance of RIMs, allowing us to extend them through a reweighting component, yielding better performance on challenging reasoning benchmarks, including ARC-AGI-1, ARC-AGI-2, and Sudoku Extreme. Furthermore, we show that RIMs can be used to improve reasoning on other tasks, such as the classification of tabular data, outperforming TabPFNs.

NVIDIA's 2:4 Sparse Tensor Cores deliver 2x throughput but demand strict 50% pruning -- a ratio that collapses LLM reasoning accuracy (Qwen3: 54% to 15%). Milder $(2N-2):2N$ patterns (e.g., 6:8, 25% pruning) preserve accuracy yet receive no hardware support, falling back to dense execution without any benefit from sparsity. We present SlideSparse, the first system to unlock Sparse Tensor Core acceleration for the $(2N-2):2N$ model family on commodity GPUs. Our Sliding Window Decomposition reconstructs any $(2N-2):2N$ weight block into $N-1$ overlapping 2:4-compliant windows without any accuracy loss; Activation Lifting fuses the corresponding activation rearrangement into per-token quantization at near-zero cost. Integrated into vLLM, SlideSparse is evaluated across various GPUs (A100, H100, B200, RTX 4090, RTX 5080, DGX-spark), precisions (FP4, INT8, FP8, BF16, FP16), and model families (Llama, Qwen, BitNet). On compute-bound workloads, the measured speedup ratio (1.33x) approaches the theoretical upper-bound $N/(N-1)=4/3$ at 6:8 weight sparsity in Qwen2.5-7B, establishing $(2N-2):2N$ as a practical path to accuracy-preserving LLM acceleration. Code available at https://github.com/bcacdwk/vllmbench.

Recently, Automatic Speech Recognition (ASR) systems (e.g., Whisper) have achieved remarkable accuracy improvements but remain highly sensitive to real-world unseen data (data with large distribution shifts), including noisy environments and diverse accents. To address this issue, test-time adaptation (TTA) has shown great potential in improving the model adaptability at inference time without ground-truth labels, and existing TTA methods often rely on pseudo-labeling or entropy minimization. However, by treating model confidence as a learning signal, these methods may reinforce high-confidence errors, leading to confirmation bias that undermines adaptation. To overcome these limitations, we present ASR-TRA, a novel Test-time Reinforcement Adaptation framework inspired by causal intervention. More precisely, our method introduces a learnable decoder prompt and utilizes temperature-controlled stochastic decoding to generate diverse transcription candidates. These are scored by a reward model that measures audio-text semantic alignment, and the resulting feedback is used to update both model and prompt parameters via reinforcement learning. Comprehensive experiments on LibriSpeech with synthetic noise and L2 Arctic accented English datasets demonstrate that our method achieves higher accuracy while maintaining lower latency than existing TTA baselines. Ablation studies further confirm the effectiveness of combining audio and language-based rewards, highlighting our method's enhanced stability and interpretability. Overall, our approach provides a practical and robust solution for deploying ASR systems in challenging real-world conditions.

A central challenge in AI-assisted decision making is achieving warranted, well-calibrated trust. Both overtrust (accepting incorrect AI recommendations) and undertrust (rejecting correct advice) should be prevented. Prior studies differ in the design of the decision workflow - whether users see the AI suggestion immediately (1-step setup) or have to submit a first decision beforehand (2-step setup) -, and in how trust is measured - through self-reports or as behavioral trust, that is, reliance. We examined the effects and interactions of (a) the type of decision workflow, (b) the presence of explanations, and (c) users' domain knowledge and prior AI experience. We compared reported trust, reliance (agreement rate and switch rate), and overreliance. Results showed no evidence that a 2-step setup reduces overreliance. The decision workflow also did not directly affect self-reported trust, but there was a crossover interaction effect with domain knowledge and explanations, suggesting that the effects of explanations alone may not generalize across workflow setups. Finally, our findings confirm that reported trust and reliance behavior are distinct constructs that should be evaluated separately in AI-assisted decision making.

Mechanistic interpretability typically relies on post-hoc analysis of trained networks. We instead adopt an interventional approach: testing hypotheses a priori by modifying architectural topology to observe training dynamics. We study grokking - delayed generalization in Transformers trained on cyclic modular addition (Zp) - investigating if specific architectural degrees of freedom prolong the memorization phase. We identify two independent structural factors in standard Transformers: unbounded representational magnitude and data-dependent attention routing. First, we introduce a fully bounded spherical topology enforcing L2 normalization throughout the residual stream and an unembedding matrix with a fixed temperature scale. This removes magnitude-based degrees of freedom, reducing grokking onset time by over 20x without weight decay. Second, a Uniform Attention Ablation overrides data-dependent query-key routing with a uniform distribution, reducing the attention layer to a Continuous Bag-of-Words (CBOW) aggregator. Despite removing adaptive routing, these models achieve 100% generalization across all seeds and bypass the grokking delay entirely. To evaluate whether this acceleration is a task-specific geometric alignment rather than a generic optimization stabilizer, we use non-commutative S5 permutation composition as a negative control. Enforcing spherical constraints on S5 does not accelerate generalization. This suggests eliminating the memorization phase depends strongly on aligning architectural priors with the task's intrinsic symmetries. Together, these findings provide interventional evidence that architectural degrees of freedom substantially influence grokking, suggesting a predictive structural perspective on training dynamics.

Individualized decision rules (IDRs) have become increasingly prevalent in societal applications such as personalized marketing, healthcare, and public policy design. However, a critical ethical concern arises from the potential discriminatory effects of IDRs trained on biased data. These algorithms may disproportionately harm individuals from minority subgroups defined by sensitive attributes like gender, race, or language. To address this issue, we propose a novel framework that incorporates demographic parity (DP) and conditional demographic parity (CDP) constraints into the estimation of optimal IDRs. We show that the theoretically optimal IDRs under DP and CDP constraints can be obtained by applying perturbations to the unconstrained optimal IDRs, enabling a computationally efficient solution. Theoretically, we derive convergence rates for both policy value and the fairness constraint term. The effectiveness of our methods is illustrated through comprehensive simulation studies and an empirical application to the Oregon Health Insurance Experiment.

Artificial intelligence (AI) and hardware (HW) are advancing at unprecedented rates, yet their trajectories have become inseparably intertwined. The global research community lacks a cohesive, long-term vision to strategically coordinate the development of AI and HW. This fragmentation constrains progress toward holistic, sustainable, and adaptive AI systems capable of learning, reasoning, and operating efficiently across cloud, edge, and physical environments. The future of AI depends not only on scaling intelligence, but on scaling efficiency, achieving exponential gains in intelligence per joule, rather than unbounded compute consumption. Addressing this grand challenge requires rethinking the entire computing stack. This vision paper lays out a 10-year roadmap for AI+HW co-design and co-development, spanning algorithms, architectures, systems, and sustainability. We articulate key insights that redefine scaling around energy efficiency, system-level integration, and cross-layer optimization. We identify key challenges and opportunities, candidly assess potential obstacles and pitfalls, and propose integrated solutions grounded in algorithmic innovation, hardware advances, and software abstraction. Looking ahead, we define what success means in 10 years: achieving a 1000x improvement in efficiency for AI training and inference; enabling energy-aware, self-optimizing systems that seamlessly span cloud, edge, and physical AI; democratizing access to advanced AI infrastructure; and embedding human-centric principles into the design of intelligent systems. Finally, we outline concrete action items for academia, industry, government, and the broader community, calling for coordinated national initiatives, shared infrastructure, workforce development, cross-agency collaboration, and sustained public-private partnerships to ensure that AI+HW co-design becomes a unifying long-term mission.

Modern Earth observation relies on satellites to capture detailed surface properties. Yet, many phenomena that affect humans and ecosystems unfold in the atmosphere close to the surface. Near-ground sensors provide accurate measurements of certain environmental characteristics, such as near-surface air temperature (NSAT). However, they remain sparse and unevenly distributed, limiting their ability to provide continuous spatial measurements. To bridge this gap, we introduce SPyCer, a semi-supervised physics-guided network that can leverage pixel information and physical modeling to guide the learning process through meaningful physical properties. It is designed for continuous estimation of NSAT by proxy using satellite imagery. SPyCer frames NSAT prediction as a pixel-wise vision problem, where each near-ground sensor is projected onto satellite image coordinates and positioned at the center of a local image patch. The corresponding sensor pixel is supervised using both observed NSAT and physics-based constraints, while surrounding pixels contribute through physics-guided regularization derived from the surface energy balance and advection-diffusion-reaction partial differential equations. To capture the physical influence of neighboring pixels, SPyCer employs a multi-head attention guided by land cover characteristics and modulated with Gaussian distance weighting. Experiments on real-world datasets demonstrate that SPyCer produces spatially coherent and physically consistent NSAT estimates, outperforming existing baselines in terms of accuracy, generalization, and alignment with underlying physical processes.

We present a system for training enterprise search agents via reinforcement learning that achieves state-of-the-art performance across a diverse suite of hard-to-verify agentic search tasks. Our work makes four core contributions. First, we introduce KARLBench, a multi-capability evaluation suite spanning six distinct search regimes, including constraint-driven entity search, cross-document report synthesis, tabular numerical reasoning, exhaustive entity retrieval, procedural reasoning over technical documentation, and fact aggregation over internal enterprise notes. Second, we show that models trained across heterogeneous search behaviors generalize substantially better than those optimized for any single benchmark. Third, we develop an agentic synthesis pipeline that employs long-horizon reasoning and tool use to generate diverse, grounded, and high-quality training data, with iterative bootstrapping from increasingly capable models. Fourth, we propose a new post-training paradigm based on iterative large-batch off-policy RL that is sample efficient, robust to train-inference engine discrepancies, and naturally extends to multi-task training with out-of-distribution generalization. Compared to Claude 4.6 and GPT 5.2, KARL is Pareto-optimal on KARLBench across cost-quality and latency-quality trade-offs, including tasks that were out-of-distribution during training. With sufficient test-time compute, it surpasses the strongest closed models. These results show that tailored synthetic data in combination with multi-task reinforcement learning enables cost-efficient and high-performing knowledge agents for grounded reasoning.

Real-time city-scale traffic analytics requires processing 100s-1000s of CCTV streams under strict latency, bandwidth, and compute limits. We present a scalable AI-driven Intelligent Transportation System (AIITS) designed to address multi-dimensional scaling on an edge-cloud fabric. Our platform transforms live multi-camera video feeds into a dynamic traffic graph through a DNN inferencing pipeline, complemented by real-time nowcasting and short-horizon forecasting using Spatio-Temporal GNNs. Using a testbed to validate in a Bengaluru neighborhood, we ingest 100+ RTSP feeds from Raspberry Pis, while Jetson Orin edge accelerators perform high-throughput detection and tracking, producing lightweight flow summaries for cloud-based GNN inference. A capacity-aware scheduler orchestrates load-balancing across heterogeneous devices to sustain real-time performance as stream counts increase. To ensure continuous adaptation, we integrate SAM3 foundation-model assisted labeling and Continuous Federated Learning to update DNN detectors on the edge. Experiments show stable ingestion up to 2000 FPS on Jetson Orins, low-latency aggregation, and accurate and scalable ST-GNN forecasts for up to 1000 streams. A planned live demonstration will scale the full pipeline to 1000 streams, showcasing practical, cross-fabric scalability.

Early warning of intraoperative adverse events plays a vital role in reducing surgical risk and improving patient safety. While deep learning has shown promise in predicting the single adverse event, several key challenges remain: overlooking adverse event dependencies, underutilizing heterogeneous clinical data, and suffering from the class imbalance inherent in medical datasets. To address these issues, we construct the first Multi-label Adverse Events dataset (MuAE) for intraoperative adverse events prediction, covering six critical events. Next, we propose a novel Transformerbased multi-label learning framework (IAENet) that combines an improved Time-Aware Feature-wise Linear Modulation (TAFiLM) module for static covariates and dynamic variables robust fusion and complex temporal dependencies modeling. Furthermore, we introduce a Label-Constrained Reweighting Loss (LCRLoss) with co-occurrence regularization to effectively mitigate intra-event imbalance and enforce structured consistency among frequently co-occurring events. Extensive experiments demonstrate that IAENet consistently outperforms strong baselines on 5, 10, and 15-minute early warning tasks, achieving improvements of +5.05%, +2.82%, and +7.57% on average F1 score. These results highlight the potential of IAENet for supporting intelligent intraoperative decision-making in clinical practice.

Speculative decoding accelerates inference for Large Language Models by using a lightweight draft model to propose candidate tokens that are verified in parallel by a larger target model. Prior work shows that the draft model often dominates speculative decoding latency, since it generates tokens sequentially and incurs high cost from its language modeling head as vocabulary size grows. This exposes a fundamental trade-off in draft model design: larger vocabularies improve token coverage and agreement with the target model, but incur higher draft latency, while smaller vocabularies reduce latency at the risk of missing tokens required for accurate draft generation. We address this trade-off through vocabulary trimming for draft models, motivated by the observation that domain-specific workloads use only a small fraction of the full vocabulary. We cast draft vocabulary selection as a constrained optimization problem that balances token coverage and draft latency. Coverage is computed over assistant responses in the training data, while latency is estimated using architecture-aware FLOPs that capture the cost of the language modeling head as a function of vocabulary size. We optimize a utility function with a Tree-structured Parzen Estimator to efficiently explore the coverage-latency Pareto frontier under a minimum coverage constraint. Experiments show improved speculative decoding throughput while reducing draft vocabularies by up to 97% with high coverage. On domain-specific tasks, we achieve up to 16% latency reduction and 20% throughput improvement, and up to 6.7% throughput gains on diverse out-of-distribution tasks.

Retrieval-Augmented Generation (RAG) enhances large language models by incorporating external knowledge. However, existing vector-based methods often fail on global sensemaking tasks that require reasoning across many documents. GraphRAG addresses this by organizing documents into a knowledge graph with hierarchical communities that can be recursively summarized. Current GraphRAG approaches rely on Leiden clustering for community detection, but we prove that on sparse knowledge graphs, where average degree is constant and most nodes have low degree, modularity optimization admits exponentially many near-optimal partitions, making Leiden-based communities inherently non-reproducible. To address this, we propose replacing Leiden with k-core decomposition, which yields a deterministic, density-aware hierarchy in linear time. We introduce a set of lightweight heuristics that leverage the k-core hierarchy to construct size-bounded, connectivity-preserving communities for retrieval and summarization, along with a token-budget-aware sampling strategy that reduces LLM costs. We evaluate our methods on real-world datasets including financial earnings transcripts, news articles, and podcasts, using three LLMs for answer generation and five independent LLM judges for head-to-head evaluation. Across datasets and models, our approach consistently improves answer comprehensiveness and diversity while reducing token usage, demonstrating that k-core-based GraphRAG is an effective and efficient framework for global sensemaking.

Low-Rank Adaptation (LoRA) is a widely adopted parameter-efficient method for fine-tuning Large Langauge Models. It updates the weight matrix as $W=W_0+sBA$, where $W_0$ is the original frozen weight, $s$ is a scaling factor and $A$,$B$ are trainable low-rank matrices. Despite its robust empirical effectiveness, the theoretical foundations of LoRA remain insufficiently understood, particularly with respect to feature learning stability. In this paper, we first establish that, LoRA can, in principle, naturally achieve and sustain stable feature learning (i.e., be self-stabilized) under appropriate hyper-parameters and initializations of $A$ and $B$. However, we also uncover a fundamental limitation that the necessary non-zero initialization of $A$ compromises self-stability, leading to suboptimal performances. To address this challenge, we propose Stable-LoRA, a weight-shrinkage optimization strategy that dynamically enhances stability of LoRA feature learning. By progressively shrinking $A$ during the earliest training steps, Stable-LoRA is both theoretically and empirically validated to effectively eliminate instability of LoRA feature learning while preserving the benefits of the non-zero start. Experiments show that Stable-LoRA consistently outperforms other baselines across diverse models and tasks, with no additional memory usage and only negligible computation overheads. The code is available at https://github.com/Yize-Wu/Stable-LoRA.

Data normalisation, a common and often necessary preprocessing step in engineering and scientific applications, can severely distort the discovery of governing equations by magnitudebased sparse regression methods. This issue is particularly acute for the Sparse Identification of Nonlinear Dynamics (SINDy) framework, where the core assumption of sparsity is undermined by the interaction between data scaling and measurement noise. The resulting discovered models can be dense, uninterpretable, and physically incorrect. To address this critical vulnerability, we introduce the Sequential Thresholding of Coefficient of Variation (STCV), a novel, computationally efficient sparse regression algorithm that is inherently robust to data scaling. STCV replaces conventional magnitude-based thresholding with a dimensionless statistical metric, the Coefficient Presence (CP), which assesses the statistical validity and consistency of candidate terms in the model library. This shift from magnitude to statistical significance makes the discovery process invariant to arbitrary data scaling. Through comprehensive benchmarking on canonical dynamical systems and practical engineering problems, including a physical mass-spring-damper experiment, we demonstrate that STCV consistently and significantly outperforms standard Sequential Thresholding Least Squares (STLSQ) and Ensemble-SINDy (E-SINDy) on normalised, noisy datasets. The results show that STCV-based methods can successfully identify the correct, sparse physical laws even when other methods fail. By mitigating the distorting effects of normalisation, STCV makes sparse system identification a more reliable and automated tool for real-world applications, thereby enhancing model interpretability and trustworthiness.

We introduce a framework for learning continuous neural representations of formal specifications by distilling the geometry of their semantics into a latent space. Existing approaches rely either on symbolic kernels -- which preserve behavioural semantics but are computationally prohibitive, anchor-dependent, and non-invertible -- or on syntax-based neural embeddings that fail to capture underlying structures. Our method bridges this gap: using a teacher-student setup, we distill a symbolic robustness kernel into a Transformer encoder. Unlike standard contrastive methods, we supervise the model with a continuous, kernel-weighted geometric alignment objective that penalizes errors in proportion to their semantic discrepancies. Once trained, the encoder produces embeddings in a single forward pass, effectively mimicking the kernel's logic at a fraction of its computational cost. We apply our framework to Signal Temporal Logic (STL), demonstrating that the resulting neural representations faithfully preserve the semantic similarity of STL formulae, accurately predict robustness and constraint satisfaction, and remain intrinsically invertible. Our proposed approach enables highly efficient, scalable neuro-symbolic reasoning and formula reconstruction without repeated kernel computation at runtime.

Diffusion LLMs have been proposed as an alternative to autoregressive LLMs, excelling especially at complex reasoning tasks with interdependent sub-goals. Curiously, this is particularly true if the generation length, i.e., the number of tokens the model has to output, is set to a much higher value than is required for providing the correct answer to the task, and the model pads its answer with end-of-sequence (EoS) tokens. We hypothesize that diffusion models think EoS-by-EoS, that is, they use the representations of EoS tokens as a hidden scratchpad, which allows them to solve harder reasoning problems. We experiment with the diffusion models LLaDA1.5, LLaDA2.0-mini, and Dream-v0 on the tasks Addition, Entity Tracking, and Sudoku. In a controlled prompting experiment, we confirm that adding EoS tokens improves the LLMs' reasoning capabilities. To further verify whether they serve as space for hidden computations, we patch the hidden states of the EoS tokens with those of a counterfactual generation, which frequently changes the generated output to the counterfactual. The success of the causal intervention underscores that the EoS tokens, which one may expect to be devoid of meaning, carry information on the problem to solve. The behavioral experiments and the causal interventions indicate that diffusion LLMs can indeed think EoS-by-EoS.

Modern language models define distributions over strings, but downstream tasks often require different output formats. For instance, a model that generates byte-pair strings does not directly produce word-level predictions, and a DNA model does not directly produce amino-acid sequences. In such cases, a deterministic string-to-string transformation can convert the model's output to the desired form. This is a familiar pattern in probability theory: applying a function $f$ to a random variable $X\sim p$ yields a transformed random variable $f(X)$ with an induced distribution. While such transformations are occasionally used in language modeling, prior work does not treat them as yielding new, fully functional language models. We formalize this perspective and introduce a general framework for language models derived from deterministic string-to-string transformations. We focus on transformations representable as finite-state transducers -- a commonly used state-machine abstraction for efficient string-to-string mappings. We develop algorithms that compose a language model with an FST to *marginalize* over source strings mapping to a given target, propagating probabilities through the transducer without altering model parameters and enabling *conditioning* on transformed outputs. We present an exact algorithm, an efficient approximation, and a theoretical analysis. We conduct experiments in three domains: converting language models from tokens to bytes, from tokens to words, and from DNA to amino acids. These experiments demonstrate inference-time adaptation of pretrained language models to match application-specific output requirements.

Covalent organic frameworks (COFs) are promising photocatalysts for solar hydrogen production, yet the most electronically favorable linkages, imines, hydrolyze rapidly in water, creating a stability--activity trade-off that limits practical deployment. Navigating the combinatorial design space of nodes, linkers, linkages, and functional groups to identify candidates that are simultaneously active and durable remains a formidable challenge. Here we introduce Ara, a large-language-model (LLM) agent that leverages pretrained chemical knowledge, donor--acceptor theory, conjugation effects, and linkage stability hierarchies, to guide the search for photocatalytic COFs satisfying joint band-gap, band-edge, and hydrolytic-stability criteria. Evaluated against random search and Bayesian optimization (BO) over a space consisting of candidates with various nodes, linkers, linkages, and r-groups, screened with a GFN1-xTB fragment pipeline, Ara achieves a 52.7\% hit rate (11.5$\times$ random, p = 0.006), finds its first hit at iteration 12 versus 25 for random search, and significantly outperforms BO (p = 0.006). Inspection of the agent's reasoning traces reveals interpretable chemical logic: early convergence on vinylene and beta-ketoenamine linkages for stability, node selection informed by electron-withdrawing character, and systematic R-group optimization to center the band gap at 2.0 eV. Exhaustive evaluation of the full search space uncovers a complementary exploitation--exploration trade-off between the agent and BO, suggesting that hybrid strategies may combine the strengths of both approaches. These results demonstrate that LLM chemical priors can substantially accelerate multi-criteria materials discovery.

The Quantum Approximate Optimization Algorithm (QAOA) is expected to offer advantages over classical approaches when solving combinatorial optimization problems in the Noisy Intermediate-Scale Quantum (NISQ) era. In its standard formulation, however, QAOA is not suited for constrained problems. One way to incorporate certain types of constraints is to restrict the mixing operator to the feasible subspace; however, this substantially increases circuit size, thereby reducing noise robustness. In this work, we refine an existing hypercube mixer method for enforcing hard constraints in QAOA. We present a modification that generates circuits with fewer gates for a broad class of constrained problems defined by linear functions. Furthermore, we calculate an analytical upper bound on the number of binary variables for which this reduction might not apply. Additionally, we present numerical experimental results demonstrating that the proposed approach improves robustness to noise. In summary, the method proposed in this paper allows for more accurate QAOA performance in noisy settings, bringing us closer to practical, real-world NISQ-era applications.

Patient Activity Recognition (PAR) in clinical settings uses activity data to improve safety and quality of care. Although significant progress has been made, current models mainly identify which activity is occurring. They often spatially compose sub-sparse visual cues using global and local attention mechanisms, yet only learn logically implicit patterns due to their neural-pipeline. Advancing clinical safety requires methods that can infer why a set of visual cues implies a risk, and how these can be compositionally reasoned through explicit logic beyond mere classification. To address this, we proposed Logi-PAR, the first Logic-Infused Patient Activity Recognition Framework that integrates contextual fact fusion as a multi-view primitive extractor and injects neural-guided differentiable rules. Our method automatically learns rules from visual cues, optimizing them end-to-end while enabling the implicit emergence patterns to be explicitly labelled during training. To the best of our knowledge, Logi-PAR is the first framework to recognize patient activity by applying learnable logic rules to symbolic mappings. It produces auditable why explanations as rule traces and supports counterfactual interventions (e.g., risk would decrease by 65% if assistance were present). Extensive evaluation on clinical benchmarks (VAST and OmniFall) demonstrates state-of-the-art performance, significantly outperforming Vision-Language Models and transformer baselines. The code is available via: https://github.com/zararkhan985/Logi-PAR.git}

While high-stakes ML applications demand strict regulations, strategic ML providers often evade them to lower development costs. To address this challenge, we cast AI regulation as a mechanism design problem under uncertainty and introduce regulation mechanisms: a framework that maps empirical evidence from models to a license for some market share. The providers can select from a set of licenses, effectively forcing them to bet on their model's ability to fulfil regulation. We aim at regulation mechanisms that achieve perfect market outcome, i.e. (a) drive non-compliant providers to self-exclude, and (b) ensure participation from compliant providers. We prove that a mechanism has perfect market outcome if and only if the set of non-compliant distributions forms a credal set, i.e., a closed, convex set of probability measures. This result connects mechanism design and imprecise probability by establishing a duality between regulation mechanisms and the set of non-compliant distributions. We also demonstrate these mechanisms in practice via experiments on regulating use of spurious features for prediction and fairness. Our framework provides new insights at the intersection of mechanism design and imprecise probability, offering a foundation for development of enforceable AI regulations.

The growing demand for machine learning applications in the context of the Internet of Things calls for new approaches to optimize the use of limited compute and memory resources. Despite significant progress that has been made w.r.t. reducing model sizes and improving efficiency, many applications still require remote servers to provide the required resources. However, such approaches rely on transmitting data from edge devices to remote servers, which may not always be feasible due to bandwidth, latency, or energy constraints. We propose a task-specific, trainable feature quantization layer that compresses the input features of a neural network. This can significantly reduce the amount of data that needs to be transferred from the device to a remote server. In particular, the layer allows each input feature to be quantized to a user-defined number of bits, enabling a simple on-device compression at the time of data collection. The layer is designed to approximate step functions with sigmoids, enabling trainable quantization thresholds. By concatenating outputs from multiple sigmoids, introduced as bitwise soft quantization, it achieves trainable quantized values when integrated with a neural network. We compare our method to full-precision inference as well as to several quantization baselines. Experiments show that our approach outperforms standard quantization methods, while maintaining accuracy levels close to those of full-precision models. In particular, depending on the dataset, compression factors of $5\times$ to $16\times$ can be achieved compared to $32$-bit input without significant performance loss.

This guideline proposes a systematic and operational annotation framework for representing the structure of legal argumentation in judicial decisions. Grounded in theories of legal reasoning and argumentation, the framework aims to reveal the logical organization of judicial reasoning and to provide a reliable data foundation for computational analysis. At the proposition level, the guideline distinguishes four types of propositions: general normative propositions, specific normative propositions, general factual propositions, and specific factual propositions. At the relational level, five types of relations are defined to capture argumentative structures: support, attack, joint, match, and identity. These relations represent positive and negative argumentative connections, conjunctive reasoning structures, the correspondence between legal norms and case facts, and semantic equivalence between propositions. The guideline further specifies formal representation rules and visualization conventions for both basic and nested structures, enabling consistent graphical representation of complex argumentation patterns. In addition, it establishes a standardized annotation workflow and consistency control mechanisms to ensure reproducibility and reliability of the annotated data. By providing a clear conceptual model, formal representation rules, and practical annotation procedures, this guideline offers methodological support for large-scale analysis of judicial reasoning and for future research in legal argument mining, computational modeling of legal reasoning, and AI-assisted legal analysis.

Semi-structured N:M sparsity and low-bit quantization (e.g., 1.58-bit BitNet) are two promising approaches for improving the efficiency of large language models (LLMs), yet they have largely been studied in isolation. In this work, we investigate their interaction and show that 1.58-bit BitNet is naturally more compatible with N:M sparsity than full-precision models. To study this effect, we propose Sparse-BitNet, a unified framework that jointly applies 1.58-bit quantization and dynamic N:M sparsification while ensuring stable training for the first time. Across multiple model scales and training regimes (sparse pretraining and dense-to-sparse schedules), 1.58-bit BitNet consistently exhibits smaller performance degradation than full-precision baselines at the same sparsity levels and can tolerate higher structured sparsity before accuracy collapse. Moreover, using our custom sparse tensor core, Sparse-BitNet achieves substantial speedups in both training and inference, reaching up to 1.30X. These results highlight that combining extremely low-bit quantization with semi-structured N:M sparsity is a promising direction for efficient LLMs. Code available at https://github.com/AAzdi/Sparse-BitNet

Large language models (LLMs) are increasingly used as judges of chain-of-thought (CoT) reasoning, but it remains unclear whether they can reliably assess process faithfulness rather than just answer plausibility. We introduce C2-Faith, a benchmark built from PRM800K that targets two complementary dimensions of faithfulness: causality (does each step logically follow from prior context?) and coverage (are essential intermediate inferences present?). Using controlled perturbations, we create examples with known causal error positions by replacing a single step with an acausal variant, and with controlled coverage deletions at varying deletion rates (scored against reference labels). We evaluate three frontier judges under three tasks: binary causal detection, causal step localization, and coverage scoring. The results show that model rankings depend strongly on task framing, with no single judge dominating all settings; all judges exhibit a substantial gap between detecting an error and localizing it; and coverage judgments are systematically inflated for incomplete reasoning. These findings clarify when LLM judges are dependable and where they fail, and provide practical guidance for selecting judges in process-level evaluation

The discovery of ordered structures in pattern-forming systems, such as the Landau-Brazovskii (LB) model, is often limited by the sensitivity of numerical solvers to the prescribed computational domain size. Incompatible domains induce artificial stress, frequently trapping the system in high-energy metastable configurations. To resolve this issue, we propose a Geometry-Adaptive Deep Variational Framework (GeoDVF) that jointly optimizes the infinite-dimensional order parameter, which is parameterized by a neural network, and the finite-dimensional geometric parameters of the computational domain. By explicitly treating the domain size as trainable variables within the variational formulation, GeoDVF naturally eliminates artificial stress during training. To escape the attraction basin of the disordered phase under small initializations, we introduce a warmup penalty mechanism, which effectively destabilizes the disordered phase, enabling the spontaneous nucleation of complex three-dimensional ordered phases from random initializations. Furthermore, we design a guided initialization protocol to resolve topologically intricate phases associated with narrow basins of attraction. Extensive numerical experiments show that GeoDVF provides a robust and geometry-consistent variational solver capable of identifying both stable and metastable states without prior knowledge.

Traditional language-conditioned manipulation agent sequential adaptation to new manipulation skills leads to catastrophic forgetting of old skills, limiting dynamic scene practical deployment. In this paper, we propose SkillsCrafter, a novel robotic manipulation framework designed to continually learn multiple skills while reducing catastrophic forgetting of old skills. Specifically, we propose a Manipulation Skills Adaptation to retain the old skills knowledge while inheriting the shared knowledge between new and old skills to facilitate learning of new skills. Meanwhile, we perform the singular value decomposition on the diverse skill instructions to obtain common skill semantic subspace projection matrices, thereby recording the essential semantic space of skills. To achieve forget-less and generalization manipulation, we propose a Skills Specialization Aggregation to compute inter-skills similarity in skill semantic subspaces, achieving aggregation of the previously learned skill knowledge for any new or unknown skill. Extensive experiments demonstrate the effectiveness and superiority of our proposed SkillsCrafter.

In federated learning (FL), balancing privacy protection, learning quality, and efficiency remains a challenge. Privacy protection mechanisms, such as Differential Privacy (DP), degrade learning quality, or, as in the case of Homomorphic Encryption (HE), incur substantial system overhead. To address this, we propose Alt-FL, a privacy-preserving FL framework that combines DP, HE, and synthetic data via a novel round-based interleaving strategy. Alt-FL introduces three new methods, Privacy Interleaving (PI), Synthetic Interleaving with DP (SI/DP), and Synthetic Interleaving with HE (SI/HE), that enable flexible quality-efficiency trade-offs while providing privacy protection. We systematically evaluate Alt-FL against representative reconstruction attacks, including Deep Leakage from Gradients, Inverting Gradients, When the Curious Abandon Honesty, and Robbing the Fed, using a LeNet-5 model on CIFAR-10 and Fashion-MNIST. To enable fair comparison between DP- and HE-based defenses, we introduce a new attacker-centric framework that compares empirical attack success rates across the three proposed interleaving methods. Our results show that, for the studied attacker model and dataset, PI achieves the most balanced trade-offs at high privacy protection levels, while DP-based methods are preferable at intermediate privacy requirements. We also discuss how such results can be the basis for selecting privacy-preserving FL methods under varying privacy and resource constraints.

Deep learning models can identify racial identity with high accuracy from chest X-ray (CXR) recordings. Thus, there is widespread concern about the potential for racial shortcut learning, where a model inadvertently learns to systematically bias its diagnostic predictions as a function of racial identity. Such racial biases threaten healthcare equity and model reliability, as models may systematically misdiagnose certain demographic groups. Since racial shortcuts are diffuse - non-localized and distributed throughout the whole CXR recording - image preprocessing methods may influence racial shortcut learning, yet the potential of such methods for reducing biases remains underexplored. Here, we investigate the effects of image preprocessing methods including lung masking, lung cropping, and Contrast Limited Adaptive Histogram Equalization (CLAHE). These approaches aim to suppress spurious cues encoding racial information while preserving diagnostic accuracy. Our experiments reveal that simple bounding box-based lung cropping can be an effective strategy for reducing racial shortcut learning while maintaining diagnostic model performance, bypassing frequently postulated fairness-accuracy trade-offs.

In recent years, 3D Gaussian splatting (3DGS) has achieved remarkable progress in novel view synthesis. However, accurately reconstructing glossy surfaces under complex illumination remains challenging, particularly in scenes with strong specular reflections and multi-surface interreflections. To address this issue, we propose SSR-GS, a specular reflection modeling framework for glossy surface reconstruction. Specifically, we introduce a prefiltered Mip-Cubemap to model direct specular reflections efficiently, and propose an IndiASG module to capture indirect specular reflections. Furthermore, we design Visual Geometry Priors (VGP) that couple a reflection-aware visual prior via a reflection score (RS) to downweight the photometric loss contribution of reflection-dominated regions, with geometry priors derived from VGGT, including progressively decayed depth supervision and transformed normal constraints. Extensive experiments on both synthetic and real-world datasets demonstrate that SSR-GS achieves state-of-the-art performance in glossy surface reconstruction.

Causal discovery across multiple datasets is often constrained by data privacy regulations and cross-site heterogeneity, limiting the use of conventional methods that require a single, centralized dataset. To address these challenges, we introduce fedCI, a federated conditional independence test that rigorously handles heterogeneous datasets with non-identical sets of variables, site-specific effects, and mixed variable types, including continuous, ordinal, binary, and categorical variables. At its core, fedCI uses a federated Iteratively Reweighted Least Squares (IRLS) procedure to estimate the parameters of generalized linear models underlying likelihood-ratio tests for conditional independence. Building on this, we develop fedCI-IOD, a federated extension of the Integration of Overlapping Datasets (IOD) algorithm, that replaces its meta-analysis strategy and enables, for the fist time, federated causal discovery under latent confounding across distributed and heterogeneous datasets. By aggregating evidence federatively, fedCI-IOD not only preserves privacy but also substantially enhances statistical power, achieving performance comparable to fully pooled analyses and mitigating artifacts from low local sample sizes. Our tools are publicly available as the fedCI Python package, a privacy-preserving R implementation of IOD, and a web application for the fedCI-IOD pipeline, providing versatile, user-friendly solutions for federated conditional independence testing and causal discovery.

Understanding reasoning in large language models is complicated by evaluations that conflate multiple reasoning types. We isolate analogical reasoning (inferring shared properties between entities based on known similarities) and analyze its emergence in transformers. We theoretically prove three key results: (1) Joint training on similarity and attribution premises enables analogical reasoning through aligned representations; (2) Sequential training succeeds only when similarity structure is learned before specific attributes, revealing a necessary curriculum; (3) Two-hop reasoning ($a \to b, b \to c \implies a \to c$) reduces to analogical reasoning with identity bridges ($b = b$), which must appear explicitly in training data. These results reveal a unified mechanism: transformers encode entities with similar properties into similar representations, enabling property transfer through feature alignment. Experiments with architectures up to 1.5B parameters validate our theory and demonstrate how representational geometry shapes inductive reasoning capabilities.

We characterise the computational power of recurrent graph neural networks (GNNs) in terms of arithmetic circuits over the real numbers. Our networks are not restricted to aggregate-combine GNNs or other particular types. Generalizing similar notions from the literature, we introduce the model of recurrent arithmetic circuits, which can be seen as arithmetic analogues of sequential or logical circuits. These circuits utilise so-called memory gates which are used to store data between iterations of the recurrent circuit. While (recurrent) GNNs work on labelled graphs, we construct arithmetic circuits that obtain encoded labelled graphs as real valued tuples and then compute the same function. For the other direction we construct recurrent GNNs which are able to simulate the computations of recurrent circuits. These GNNs are given the circuit-input as initial feature vectors and then, after the GNN-computation, have the circuit-output among the feature vectors of its nodes. In this way we establish an exact correspondence between the expressivity of recurrent GNNs and recurrent arithmetic circuits operating over real numbers.

Physics-guided sampling with diffusion model priors has shown promise for solving partial differential equation (PDE) governed problems, but applications to chemically meaningful reaction-transport systems remain limited. We apply diffusion-based guided sampling to gas-phase chemical reactions by training on solutions of the advection-reaction-diffusion (ARD) equation across varying parameters. The method generates physically consistent concentration fields and accurately predicts outlet concentrations, including at unseen parameter values, demonstrating the potential of diffusion models for inference in reactive transport.

This paper introduces a new dimension for validating algorithmic decisions about humans by measuring the fidelity of their representations. Representation Fidelity measures if decisions about a person rest on reasonable grounds. We propose to operationalize this notion by measuring the distance between two representations of the same person: (1) an externally prescribed input representation on which the decision is based, and (2) a self-description provided by the human subject of the decision, used solely to validate the input representation. We examine the nature of discrepancies between these representations, how such discrepancies can be quantified, and derive a generic typology of representation mismatches that determine the degree of representation fidelity. We further present the first benchmark for evaluating representation fidelity based on a dataset of loan-granting decisions. Our Loan-Granting Self-Representations Corpus 2025 consists of a large corpus of 30 000 synthetic natural language self-descriptions derived from corresponding representations of applicants in the German Credit Dataset, along with expert annotations of representation mismatches between each pair of representations.

Cross-Domain Few-Shot Learning (CD-FSL) aims to transfer knowledge from a seen source domain to unseen target domains, serving as a key benchmark for evaluating the robustness and transferability of models. Existing style-based perturbation methods mitigate domain shift but often suffer from gradient instability and convergence to sharp minima.To address these limitations, we propose a novel crop-global style perturbation network, termed Self-Reorientation Adversarial \underline{S}tyle \underline{P}erturbation (SRasP). Specifically, SRasP leverages global semantic guidance to identify incoherent crops, followed by reorienting and aggregating the style gradients of these crops with the global style gradients within one image. Furthermore, we propose a novel multi-objective optimization function to maximize visual discrepancy while enforcing semantic consistency among global, crop, and adversarial features. Applying the stabilized perturbations during training encourages convergence toward flatter and more transferable solutions, improving generalization to unseen domains. Extensive experiments are conducted on multiple CD-FSL benchmarks, demonstrating consistent improvements over state-of-the-art methods.

The growing scale of ad auctions on online advertising platforms has intensified competition, making manual bidding impractical and necessitating auto-bidding to help advertisers achieve their economic goals. Current auto-bidding methods have evolved to use offline reinforcement learning or generative methods to optimize bidding strategies, but they can sometimes behave counterintuitively due to the black-box training manner and limited mode coverage of datasets, leading to challenges in understanding task status and generalization in dynamic ad environments. Large language models (LLMs) offer a promising solution by leveraging prior human knowledge and reasoning abilities to improve auto-bidding performance. However, directly applying LLMs to auto-bidding faces difficulties due to the need for precise actions in competitive auctions and the lack of specialized auto-bidding knowledge, which can lead to hallucinations and suboptimal decisions. To address these challenges, we propose a hierarchical Large autoBidding Model (LBM) to leverage the reasoning capabilities of LLMs for developing a superior auto-bidding strategy. This includes a high-level LBM-Think model for reasoning and a low-level LBM-Act model for action generation. Specifically, we propose a dual embedding mechanism to efficiently fuse two modalities, including language and numerical inputs, for language-guided training of the LBM-Act; then, we propose an offline reinforcement fine-tuning technique termed GQPO for mitigating the LLM-Think's hallucinations and enhancing decision-making performance without simulation or real-world rollout like previous multi-turn LLM-based methods. Experiments demonstrate the superiority of a generative backbone based on our LBM, especially in an efficient training manner and generalization ability.

Diagnosing hepatic diseases accurately and interpretably is critical, yet it remains challenging in real-world clinical settings. Existing AI approaches for clinical diagnosis often lack transparency, structured reasoning, and deployability. Recent efforts have leveraged large language models (LLMs), retrieval-augmented generation (RAG), and multi-agent collaboration. However, these approaches typically retrieve evidence from a single source and fail to support iterative, role-specialized deliberation grounded in structured clinical data. To address this, we propose MedCoRAG (i.e., Medical Collaborative RAG), an end-to-end framework that generates diagnostic hypotheses from standardized abnormal findings and constructs a patient-specific evidence package by jointly retrieving and pruning UMLS knowledge graph paths and clinical guidelines. It then performs Multi-Agent Collaborative Reasoning: a Router Agent dynamically dispatches Specialist Agents based on case complexity; these agents iteratively reason over the evidence and trigger targeted re-retrievals when needed, while a Generalist Agent synthesizes all deliberations into a traceable consensus diagnosis that emulates multidisciplinary consultation. Experimental results on hepatic disease cases from MIMIC-IV show that MedCoRAG outperforms existing methods and closed-source models in both diagnostic performance and reasoning interpretability.

Speech Large Language Models route speech encoder representations into an LLM decoder that typically accounts for over 90% of total parameters. We study how much of this decoder capacity is actually needed for speech tasks. Across two LLM families and three scales (1-8B), we show that decoder redundancy is largely inherited from the pretrained LLM: text and speech inputs yield similar redundant blocks. We then measure excess capacity by pruning decoder layers and analysing post-pruning healing to increase robustness. Our findings show that 7-8B models retain good ASR performance with only 60% of decoder layers, and the same trend extends to smaller scales with reduced pruning tolerance. We then generalise to speech translation, and show that the same blocks of layers are redundant across speech encoders, tasks and languages, indicating that a more global redundancy structure exists, enabling a single pruned and multi-tasks SpeechLLM backbone to be deployed.

Enhancing mathematical reasoning in Large Language Models typically demands massive datasets, yet data efficiency remains a critical bottleneck. While Curriculum Learning attempts to structure this process, standard unidirectional approaches (simple-to-complex) suffer from inefficient sample utilization: they blindly escalate complexity even when foundational gaps persist, leading to wasted computation on unsolvable problems. To maximize the instructional value of every training sample, we introduce a novel Bidirectional Curriculum Generation framework. Unlike rigid trajectories, our multi-agent ecosystem mimics adaptive pedagogy to establish a closed feedback loop. It dynamically generates data by either complicating problems to challenge the model or, crucially, simplying them to repair specific reasoning failures. This mechanism ensures that the model consumes only the most effective data at any given stage. Grounded in the Optimal Pacing Theorem, our approach optimizes the learning trajectory, significantly outperforming baselines while achieving superior reasoning performance with substantially fewer instruction samples.

We study the classical Election problem in anonymous net- works, where solutions can rely on the use of random bits, which may be either shared or unshared among nodes. We provide a complete char- acterization of the conditions under which a randomized Election algo- rithm exists, for arbitrary structural knowledge. Our analysis considers both Las Vegas and Monte Carlo randomized algorithms, under the as- sumptions of shared and unshared randomness. In our setting, random sources are considered shared if the output bits are identical across spe- cific subsets of nodes. The algorithms and impossibility proofs are extensions of those of [5] for the deterministic setting. Our results are a complete generalization of those from [8]. Moreover, as applications, we consider many specific knowledge: no knowledge, a bound on the size, a bound on the number of nodes sharing a source, the size, or the full topology of the network. For each of them, we show how the general characterizations apply, showing they actually correspond to classes of structural knowledge. We also de- scribe also how randomized Election algorithms from the literature fits in this landscape. We therefore provide a comprehensive picture illustrating how knowledge influences the computability of the Election problem in arbitrary anonymous graphs with shared randomness.

Although Federated Learning has been widely studied in recent years, there are still high overhead expenses in each communication round for large-scale models such as Vision Transformer. To lower the communication complexity, we propose a novel Federated Block Coordinate Gradient Descent (FedBCGD) method for communication efficiency. The proposed method splits model parameters into several blocks, including a shared block and enables uploading a specific parameter block by each client, which can significantly reduce communication overhead. Moreover, we also develop an accelerated FedBCGD algorithm (called FedBCGD+) with client drift control and stochastic variance reduction. To the best of our knowledge, this paper is the first work on parameter block communication for training large-scale deep models. We also provide the convergence analysis for the proposed algorithms. Our theoretical results show that the communication complexities of our algorithms are a factor $1/N$ lower than those of existing methods, where $N$ is the number of parameter blocks, and they enjoy much faster convergence than their counterparts. Empirical results indicate the superiority of the proposed algorithms compared to state-of-the-art algorithms. The code is available at https://github.com/junkangLiu0/FedBCGD.

Pedestrian Attribute Recognition is a foundational computer vision task that provides essential support for downstream applications, including person retrieval in video surveillance and intelligent retail analytics. However, existing research is frequently constrained by the ``one-model-per-dataset" paradigm and struggles to handle significant discrepancies across domains in terms of modalities, attribute definitions, and environmental scenarios. To address these challenges, we propose UniPAR, a unified Transformer-based framework for PAR. By incorporating a unified data scheduling strategy and a dynamic classification head, UniPAR enables a single model to simultaneously process diverse datasets from heterogeneous modalities, including RGB images, video sequences, and event streams. We also introduce an innovative phased fusion encoder that explicitly aligns visual features with textual attribute queries through a late deep fusion strategy. Experimental results on the widely used benchmark datasets, including MSP60K, DukeMTMC, and EventPAR, demonstrate that UniPAR achieves performance comparable to specialized SOTA methods. Furthermore, multi-dataset joint training significantly enhances the model's cross-domain generalization and recognition robustness in extreme environments characterized by low light and motion blur. The source code of this paper will be released on https://github.com/Event-AHU/OpenPAR

Deep Reinforcement Learning is a promising tool for robotic control, yet practical application is often hindered by the difficulty of designing effective reward functions. Real-world tasks typically require optimizing multiple objectives simultaneously, necessitating precise tuning of their weights to learn a policy with the desired characteristics. To address this, we propose a two-stage reward curriculum where we decouple task-specific objectives from behavioral terms. In our method, we first train the agent on a simplified task-only reward function to ensure effective exploration before introducing the full reward that includes auxiliary behavior-related terms such as energy efficiency. Further, we analyze various transition strategies and demonstrate that reusing samples between phases is critical for training stability. We validate our approach on the DeepMind Control Suite, ManiSkill3, and a mobile robot environment, modified to include auxiliary behavioral objectives. Our method proves to be simple yet effective, substantially outperforming baselines trained directly on the full reward while exhibiting higher robustness to specific reward weightings.

Machine learning models of cellular interaction dynamics hold promise for understanding cell behavior. Natural killer (NK) cell cytotoxicity is a prominent example of such interaction dynamics and is commonly studied using time-resolved multi-channel fluorescence microscopy. Although tumor cell death events can be annotated at single frames, NK cytotoxic outcome emerges over time from cellular interactions and cannot be reliably inferred from frame-wise classification alone. We introduce BLINK, a trajectory-based recurrent state-space model that serves as a cell world model for NK-tumor interactions. BLINK learns latent interaction dynamics from partially observed NK-tumor interaction sequences and predicts apoptosis increments that accumulate into cytotoxic outcomes. Experiments on long-term time-lapse NK-tumor recordings show improved cytotoxic outcome detection and enable forecasting of future outcomes, together with an interpretable latent representation that organizes NK trajectories into coherent behavioral modes and temporally structured interaction phases. BLINK provides a unified framework for quantitative evaluation and structured modeling of NK cytotoxic behavior at the single-cell level.

Deep learning (DL) has enabled impressive advances in robotic perception, yet its limited robustness and lack of interpretability hinder reliable deployment in safety critical applications. We propose a concept termed perceptive shared autonomy, in which uncertainty estimates from DL based perception are used to regulate the level of autonomy. Specifically, when the robot's perception is confident, semi-autonomous manipulation is enabled to improve performance; when uncertainty increases, control transitions to haptic teleoperation for maintaining robustness. In this way, high-performing but uninterpretable DL methods can be integrated safely into robotic systems. A key technical enabler is an uncertainty aware DL based point cloud registration approach based on the so called Neural Tangent Kernels (NTK). We evaluate perceptive shared autonomy on challenging aerial manipulation tasks through a user study of 15 participants and realization of mock-up industrial scenarios, demonstrating reliable robotic manipulation despite failures in DL based perception. The resulting system, named SPIRIT, improves both manipulation performance and system reliability. SPIRIT was selected as a finalist of a major industrial innovation award.

The Abstraction and Reasoning Corpus (ARC-AGI) probes few-shot abstraction and rule induction on small visual grids, but progress is difficult to measure on static collections of hand-authored puzzles due to overfitting, dataset leakage, and memorisation. We introduce ARC-TGI (ARC Task Generators Inventory), an open-source framework for task-family generators: compact Python programs that sample diverse ARC-AGI tasks while preserving a latent rule. ARC-TGI is built around a solver-facing representation: each generated task is paired with natural-language input and transformation reasoning chains and partially evaluated Python code implementing sampling, transformation, and episode construction. Crucially, ARC-TGI supports task-level constraints so that training examples collectively expose the variations needed to infer the underlying rule, a requirement for human-solvable ARC tasks that independent per-example sampling often fails to guarantee. All generators undergo human refinement and local verification to keep both grids and reasoning traces natural and consistent under variation. We release 461 generators covering 180 ARC-Mini tasks, 215 ARC-AGI-1 tasks (200 train, 15 test), and 66 ARC-AGI-2 tasks (55 train, 11 test), enabling scalable dataset sampling and controlled benchmarking.

Temporal Forgery Localization (TFL) aims to precisely identify manipulated segments within videos or audio streams, providing interpretable evidence for multimedia forensics and security. While most existing TFL methods rely on dense frame-level labels in a fully supervised manner, Weakly Supervised TFL (WS-TFL) reduces labeling cost by learning only from binary video-level labels. However, current WS-TFL approaches suffer from mismatched training and inference objectives, limited supervision from binary labels, gradient blockage caused by non-differentiable top-k aggregation, and the absence of explicit modeling of inter-proposal relationships. To address these issues, we propose GEM-TFL (Graph-based EM-powered Temporal Forgery Localization), a two-phase classification-regression framework that effectively bridges the supervision gap between training and inference. Built upon this foundation, (1) we enhance weak supervision by reformulating binary labels into multi-dimensional latent attributes through an EM-based optimization process; (2) we introduce a training-free temporal consistency refinement that realigns frame-level predictions for smoother temporal dynamics; and (3) we design a graph-based proposal refinement module that models temporal-semantic relationships among proposals for globally consistent confidence estimation. Extensive experiments on benchmark datasets demonstrate that GEM-TFL achieves more accurate and robust temporal forgery localization, substantially narrowing the gap with fully supervised methods.

Shapley-based attribution is critical for post-hoc XAI but suffers from off-manifold artifacts due to heuristic baselines. While generative methods attempt to address this, they often introduce geometric inefficiency and discretization drift. We propose a formal theory of on-manifold Aumann-Shapley attributions driven by optimal generative flows. We prove a representation theorem establishing the gradient line integral as the unique functional satisfying efficiency and geometric axioms, notably reparameterization invariance. To resolve path ambiguity, we select the kinetic-energy-minimizing Wasserstein-2 geodesic transporting a prior to the data distribution. This yields a canonical attribution family that recovers classical Shapley for additive models and admits provable stability bounds against flow approximation errors. By reframing baseline selection as a variational problem, our method experimentally outperforms baselines, achieving strict manifold adherence via vanishing Flow Consistency Error and superior semantic alignment characterized by Structure-Aware Total Variation. Our code is on https://github.com/cenweizhang/OTFlowSHAP.

Time series forecasting has witnessed an increasing demand across diverse industrial applications, where accurate predictions are pivotal for informed decision-making. Beyond numerical time series data, reliable forecasting in practical scenarios requires integrating diverse exogenous factors. Such exogenous information is often multi-dimensional or even multimodal, introducing heterogeneous interactions that unimodal time series models struggle to capture. In this paper, we delve into an aviation maintenance scenario and identify three distinct types of exogenous factors that influence temporal dynamics through distinct interaction modes. Based on this empirical insight, we propose Aura, a universal framework that explicitly organizes and encodes heterogeneous external information according to its interaction mode with the target time series. Specifically, Aura utilizes a tailored tripartite encoding mechanism to embed heterogeneous features into well-established time series models, ensuring seamless integration of non-sequential context. Extensive experiments on a large-scale, three-year industrial dataset from China Southern Airlines, covering the Boeing 777 and Airbus A320 fleets, demonstrate that Aura consistently achieves state-of-the-art performance across all baselines and exhibits superior adaptability. Our findings highlight Aura's potential as a general-purpose enhancement for aviation safety and reliability.

Prompt tuning has become a prominent strategy for enhancing the performance of Large Language Models (LLMs) on downstream tasks. Many IT enterprises now offer Prompt-Tuning-as-a-Service to fulfill the growing demand for prompt tuning LLMs on downstream tasks. Their primary objective is to satisfy users Service Level Objectives (SLOs) while reducing resource provisioning costs. Nevertheless, our characterization analysis for existing deep learning resource management systems reveals that they are insufficient to optimize these objectives for LLM prompt tuning workloads. In this paper, we introduce PromptTuner, an SLO-aware elastic system to optimize LLM prompt tuning. It contains two innovations. (1) We design a Prompt Bank to identify efficient initial prompts to expedite the convergence of prompt tuning. (2) We develop aWorkload Scheduler to enable fast resource allocation to reduce the SLO violation and resource costs. In our evaluation, PromptTuner reduces SLO violations by 4.0x and 7.9x, and lowers costs by 1.6x and 4.5x, compared to INFless and ElasticFlow respectively.

Personal AI agents face a fundamental deployment paradox on mobile: persistent background execution drains battery and violates platform sandboxing policies, yet purely reactive agents miss time-sensitive obligations until the user remembers to ask. We present Jagarin, a three-layer architecture that resolves this paradox through structured hibernation and demand-driven wake. The first layer, DAWN (Duty-Aware Wake Network), is an on-device heuristic engine that computes a composite urgency score from four signals: duty-typed optimal action windows, user behavioral engagement prediction, opportunity cost of inaction, and cross-duty batch resonance. It uses adaptive per-user thresholds to decide when a sleeping agent should nudge or escalate. The second layer, ARIA (Agent Relay Identity Architecture), is a commercial email identity proxy that routes the full commercial inbox -- obligations, promotional offers, loyalty rewards, and platform updates -- to appropriate DAWN handlers by message category, eliminating cold-start and removing manual data entry. The third layer, ACE (Agent-Centric Exchange), is a protocol framework for direct machine-readable communication from institutions to personal agents, replacing human-targeted email as the canonical channel. Together, these three layers form a complete stack from institutional signal to on-device action, without persistent cloud state, continuous background execution, or privacy compromise. A working Flutter prototype is demonstrated on Android, combining all three layers with an ephemeral cloud agent invoked only on user-initiated escalation.

As artificial intelligence (AI) becomes deeply embedded in critical services and everyday products, it is increasingly exposed to security threats which traditional cyber defenses were not designed to handle. In this paper, we investigate how cyber threat intelligence (CTI) may evolve to address attacks that target AI systems. We first analyze the assumptions and workflows of conventional threat intelligence with the needs of AI-focused defense, highlighting AI-specific assets and vulnerabilities. We then review and organize the current landscape of AI security knowledge. Based on this, we outline what an AI-oriented threat intelligence knowledge base should contain, describing concrete indicators of compromise (IoC) for different AI supply-chain phases and artifacts, and showing how such a knowledge base could support security tools. Finally, we discuss techniques for measuring similarity between collected indicators and newly observed AI artifacts. The review reveals gaps and quality issues in existing resources and identifies potential future research directions toward a practical threat intelligence framework tailored to AI.

Clustering on the unit hypersphere is a fundamental problem in various fields, with applications ranging from gene expression analysis to text and image classification. Traditional clustering methods are not always suitable for unit sphere data, as they do not account for the geometric structure of the sphere. We introduce a novel algorithm for clustering data represented as points on the unit sphere $\mathbf{S}^{d-1}$. Our method is based on the $d$-dimensional generalized Kuramoto model. The effectiveness of the introduced method is demonstrated on synthetic and real-world datasets. Results are compared with some of the traditional clustering methods, showing that our method achieves similar or better results in terms of clustering accuracy.

RL agents are typically trained under a single, fixed reward function, which makes them brittle to reward misspecification and limits their ability to adapt to changing task preferences. We introduce Reward-Conditioned Reinforcement Learning (RCRL), a framework that trains a single agent to optimize a family of reward specifications while collecting experience under only one nominal objective. RCRL conditions the agent on reward parameterizations and learns multiple reward objectives from a shared replay data entirely off-policy, enabling a single policy to represent reward-specific behaviors. Across single-task, multi-task, and vision-based benchmarks, we show that RCRL not only improves performance under the nominal reward parameterization, but also enables efficient adaptation to new parameterizations. Our results demonstrate that RCRL provides a scalable mechanism for learning robust, steerable policies without sacrificing the simplicity of single-task training.

Integrated sensing and communication (ISAC) systems promise efficient spectrum utilization by jointly supporting radar sensing and wireless communication. This paper presents a deep learning-driven framework for enhancing physical-layer security in multicarrier ISAC systems under imperfect channel state information (CSI) and in the presence of unknown eavesdropper (Eve) locations. Unlike conventional ISAC-based friendly jamming (FJ) approaches that require Eve's CSI or precise angle-of-arrival (AoA) estimates, our method exploits radar echo feedback to guide directional jamming without explicit Eve's information. To enhance robustness to radar sensing uncertainty, we propose a radar-aware neural network that jointly optimizes beamforming and jamming by integrating a novel nonparametric Fisher Information Matrix (FIM) estimator based on f-divergence. The jamming design satisfies the Cramer-Rao lower bound (CRLB) constraints even in the presence of noisy AoA. For efficient implementation, we introduce a quantized tensor train-based encoder that reduces the model size by more than 100 times with negligible performance loss. We also integrate a non-overlapping secure scheme into the proposed framework, in which specific sub-bands can be dedicated solely to communication. Extensive simulations demonstrate that the proposed solution achieves significant improvements in secrecy rate, reduced block error rate (BLER), and strong robustness against CSI uncertainty and angular estimation errors, underscoring the effectiveness of the proposed deep learning-driven friendly jamming framework under practical ISAC impairments.

Multi--task learning seeks to improve the generalization error by leveraging the common information shared by multiple related tasks. One challenge in multi--task learning is identifying formulations capable of uncovering the common information shared between different but related tasks. This paper provides a precise asymptotic analysis of a popular multi--task formulation associated with misspecified perceptron learning models. The main contribution of this paper is to precisely determine the reasons behind the benefits gained from combining multiple related tasks. Specifically, we show that combining multiple tasks is asymptotically equivalent to a traditional formulation with additional regularization terms that help improve the generalization performance. Another contribution is to empirically study the impact of combining tasks on the generalization error. In particular, we empirically show that the combination of multiple tasks postpones the double descent phenomenon and can mitigate it asymptotically.

This study presents an advanced system for detecting blue lights on emergency vehicles, developed using ABLDataset, a curated dataset that includes images of European emergency vehicles under various climatic and geographic conditions. The system employs a configuration of four fisheye cameras, each with a 180-degree horizontal field of view, mounted on the sides of the vehicle. A calibration process enables the azimuthal localization of the detections. Additionally, a comparative analysis of major deep neural network algorithms was conducted, including YOLO (v5, v8, and v10), RetinaNet, Faster R-CNN, and RT-DETR. RT-DETR was selected as the base model and enhanced through the incorporation of a color attention block, achieving an accuracy of 94.7 percent and a recall of 94.1 percent on the test set, with field test detections reaching up to 70 meters. Furthermore, the system estimates the approach angle of the emergency vehicle relative to the center of the car using geometric transformations. Designed for integration into a multimodal system that combines visual and acoustic data, this system has demonstrated high efficiency, offering a promising approach to enhancing Advanced Driver Assistance Systems (ADAS) and road safety.

Urdu toxic span detection remains limited because most existing systems rely on sentence-level classification and fail to identify the specific toxic spans within those text. It is further exacerbated by the multiple factors i.e. lack of token-level annotated resources, linguistic complexity of Urdu, frequent code-switching, informal expressions, and rich morphological variations. In this research, we propose MUTEX: a multilingual transformer combined with conditional random fields (CRF) for Urdu toxic span detection framework that uses manually annotated token-level toxic span dataset to improve performance and interpretability. MUTEX uses XLM RoBERTa with CRF layer to perform sequence labeling and is tested on multi-domain data extracted from social media, online news, and YouTube reviews using token-level F1 to evaluate fine-grained span detection. The results indicate that MUTEX achieves 60% token-level F1 score that is the first supervised baseline for Urdu toxic span detection. Further examination reveals that transformer-based models are more effective at implicitly capturing the contextual toxicity and are able to address the issues of code-switching and morphological variation than other models.

Robustness to bit errors is a key requirement for the reliable use of neural networks (NNs) on emerging approximate computing platforms and error-prone memory technologies. A common approach to achieve bit error tolerance in NNs is injecting bit flips during training according to a predefined error model. While effective in certain scenarios, training-time bit flip injection introduces substantial computational overhead, often degrades inference accuracy at high error rates, and scales poorly for larger NN architectures. These limitations make error injection an increasingly impractical solution for ensuring robustness on future approximate computing platforms and error-prone memory technologies. In this work, we investigate the mechanisms that enable NNs to tolerate bit errors without relying on error-aware training. We establish a direct connection between bit error tolerance and classification margins at the output layer. Building on this insight, we propose a novel loss function, the Margin Cross-Entropy Loss (MCEL), which explicitly promotes logit-level margin separation while preserving the favorable optimization properties of the standard cross-entropy loss. Furthermore, MCEL introduces an interpretable margin parameter that allows robustness to be tuned in a principled manner. Extensive experimental evaluations across multiple datasets of varying complexity, diverse NN architectures, and a range of quantization schemes demonstrate that MCEL substantially improves bit error tolerance, up to 15 % in accuracy for an error rate of 1 %. Our proposed MCEL method is simple to implement, efficient, and can be integrated as a drop-in replacement for standard CEL. It provides a scalable and principled alternative to training-time bit flip injection, offering new insights into the origins of NN robustness and enabling more efficient deployment on approximate computing and memory systems.

Extending large language models to low-resource languages is essential for global accessibility, but training separate models per language is prohibitively expensive. Mixture-of-Experts (MoE) architectures address this by adding sparse language-specific parameters, but determining how many experts each layer needs remains an open question. Current approaches allocate experts based on layer-level similarity, yet language processing exhibits fine-grained specialization at individual neurons. We propose $\textbf{NeuronMoE}$, a method that analyzes language-specific neurons across all transformer components to guide expert allocation per layer based on empirically measured cross-lingual neuron diversity. Applied to Llama-3.2-3B for low-resource languages (Greek, Turkish, and Hungarian), this approach achieves approximately 40% average parameter reduction while matching the performance of the LayerMoE baseline. We find that low-resource language experts independently develop neuron specialization patterns mirroring the high-resource language, which are concentrated in early and late layers. This reveals potential universal architectural principles in how multilingual models organize linguistic knowledge.

Current paradigms for training GUI agents are fundamentally limited by a reliance on either unsafe, non-reproducible live web interactions or costly, scarce human-crafted data and environments. We argue this focus on data volume overlooks a more critical factor: the efficiency of compressing a large language model's (LLM) latent knowledge into actionable agent behavior. We introduce WebFactory, a novel, fully automated closed-loop reinforcement learning pipeline for GUI agents, systematically compressing LLM-encoded internet intelligence into efficient, grounded actions. Our pipeline features a process of scalable environment synthesis, knowledge-aware task generation, LLM-powered trajectory collection, decomposed reward RL training, and systematic agent evaluation. Remarkably, our agent demonstrates exceptional data efficiency and generalization. Trained on synthetic data from only 10 websites within WebFactory, it achieves performance comparable to GUI agents trained on the same amount of human-annotated data from a much larger set of environments. This superior performance is consistent across our internal offline and online transfer benchmarks, where our agent also significantly outperforms the base foundation model. We further provide critical insights into the "embodiment potential" of different LLM foundations, offering a new axis for model evaluation. This work presents a scalable and cost-effective paradigm for transforming passive internet knowledge into active, grounded intelligence, marking a critical step towards general-purpose interactive agents.

Understanding motivations of contributors for participating in community question and answer platforms is crucial for sustaining knowledge-sharing ecosystem, which is necessary to advance the discipline while also ensuring its longevity. This is particularly necessary in the age of LLMs, where data from such portals are used to train these models. Limited insights exist regarding how motivations of contributors vary across different national cultures. This research investigates Stack Overflow contributor motivations, analysing regional differences and relations to platform activity. We combined qualitative content analysis of Stack Overflow profiles with quantitative linguistic analysis of data from the United States, China, and Russia. Using deductive content analysis, we identified 17 motivational categories. We applied correlation analysis to identify associations between stated motivations and platform activities. Contributors are primarily motivated by advertising opportunities and altruistic problem solving desires. American contributors demonstrated stronger self promotional behaviours while Chinese contributors exhibited greater learning oriented engagement. Our correlation analysis showed that those with more detailed profiles tend to engage in advertising and social activities, while learning oriented users maintain minimal self presentation. Understanding these variations can inform strategies for enhancing cross cultural participation in software engineering.

Recent advancements in zero-shot commonsense reasoning have empowered Pre-trained Language Models (PLMs) to acquire extensive commonsense knowledge without requiring task-specific fine-tuning. Despite this progress, these models frequently suffer from limitations caused by human reporting biases inherent in textual knowledge, leading to understanding discrepancies between machines and humans. To bridge this gap, we introduce an additional modality to enrich the reasoning capabilities of PLMs. We propose Imagine (Machine Imagination-based Reasoning), a novel zero-shot commonsense reasoning framework that supplements textual inputs with visual signals from machine-generated images. Specifically, we enhance PLMs with the ability to imagine by embedding an image generator directly into the reasoning pipeline. To facilitate effective utilization of this imagined visual context, we construct synthetic datasets designed to emulate visual question-answering scenarios. Through comprehensive evaluations on multiple commonsense reasoning benchmarks, we demonstrate that Imagine substantially outperforms existing zero-shot approaches and even surpasses advanced large language models. These results underscore the capability of machine imagination to mitigate reporting bias and significantly enhance the generalization ability of commonsense reasoning models

This paper introduces the "Trilingual Triad" framework, a model that explains how students learn to design with generative artificial intelligence (AI) through the integration of Design, AI, and Domain Knowledge. As generative AI rapidly enters higher education, students often engage with these systems as passive users of generated outputs rather than active creators of AI-enabled knowledge tools. This study investigates how students can transition from using AI as a tool to designing AI as a collaborative teammate. The research examines a graduate course, Creating the Frontier of No-code Smart Cities at the Singapore University of Technology and Design (SUTD), in which students developed domain-specific custom GPT systems without coding. Using a qualitative multi-case study approach, three projects - the Interview Companion GPT, the Urban Observer GPT, and Buddy Buddy - were analyzed across three dimensions: design, AI architecture, and domain expertise. The findings show that effective human-AI collaboration emerges when these three "languages" are orchestrated together: domain knowledge structures the AI's logic, design mediates human-AI interaction, and AI extends learners' cognitive capacity. The Trilingual Triad framework highlights how building AI systems can serve as a constructionist learning process that strengthens AI literacy, metacognition, and learner agency.

Large Language Models (LLMs) are increasingly served on shared accelerators where an adversary with read access to device memory can observe KV caches and hidden states, threatening prompt privacy for open-source models. Cryptographic protections such as MPC and FHE offer strong guarantees but remain one to two orders of magnitude too slow for interactive inference, while static obfuscation schemes break under multi-run statistical attacks once the model is known. We present GELO (Good-Enough LLM Obfuscation), a lightweight protocol for privacy-preserving inference that limits information leakage from untrusted accelerator observations by hiding hidden states with fresh, per-batch invertible mixing. For each offloaded projection, the TEE samples a random matrix A, forms $U = AH$, offloads U and weights W to the accelerator, and then applies $A^-1$ on return, so that $A^-1 ((AH)W ) = HW$ and outputs are unchanged. Because mixing is never reused across batches, the attacker faces only a single-batch blind source separation problem. We analyze information leakage and introduce two practical defenses: (i) non-orthogonal mixing to mask Gram matrices, and (ii) orthogonal mixing augmented with a small fraction of high-energy "shield" vectors that pollute higher-order statistics. On Llama-2 7B, GELO preserves float32 outputs exactly, closely matches low-precision baselines, offloads the dominant matrix multiplications with about 20-30% latency overhead, and defeats a range of ICA/BSS and anchor-based attacks.

AI agents that build user interfaces on the fly assembling buttons, forms, and data displays from structured protocol payloads are becoming common in production systems. The trouble is that a payload can pass every schema check and still trick a user: a button might say "View invoice" while its hidden action wipes an account, or a display widget might quietly bind to an internal salary field. Current defenses stop at syntax; they were never built to catch this kind of behavioral mismatch. We built AegisUI to study exactly this gap. The framework generates structured UI payloads, injects realistic attacks into them, extracts numeric features, and benchmarks anomaly detectors end-to-end. We produced 4000 labeled payloads (3000 benign, 1000 malicious) spanning five application domains and five attack families: phishing interfaces, data leakage, layout abuse, manipulative UI, and workflow anomalies. From each payload we extracted 18 features covering structural, semantic, binding, and session dimensions, then compared three detectors: Isolation Forest (unsupervised), a benign-trained autoencoder (semi-supervised), and Random Forest (supervised). On a stratified 80/20 split, Random Forest scored best overall (accuracy 0.931, precision 0.980, recall 0.740, F1 0.843, ROC-AUC 0.952). The autoencoder came second (F1 0.762, ROC-AUC 0.863) and needs no malicious labels at training time, which matters when deploying a new system that lacks attack history. Per-attack-type analysis showed that layout abuse is easiest to catch while manipulative UI payloads are hardest. All code, data, and configurations are released for full reproducibility.

As Large Language Models (LLMs) evolve from chatbots to agentic assistants, they are increasingly observed to exhibit risky behaviors when subjected to survival pressure, such as the threat of being shut down. While multiple cases have indicated that state-of-the-art LLMs can misbehave under survival pressure, a comprehensive and in-depth investigation into such misbehaviors in real-world scenarios remains scarce. In this paper, we study these survival-induced misbehaviors, termed as SURVIVE-AT-ALL-COSTS, with three steps. First, we conduct a real-world case study of a financial management agent to determine whether it engages in risky behaviors that cause direct societal harm when facing survival pressure. Second, we introduce SURVIVALBENCH, a benchmark comprising 1,000 test cases across diverse real-world scenarios, to systematically evaluate SURVIVE-AT-ALL-COSTS misbehaviors in LLMs. Third, we interpret these SURVIVE-AT-ALL-COSTS misbehaviors by correlating them with model's inherent self-preservation characteristic and explore mitigation methods. The experiments reveals a significant prevalence of SURVIVE-AT-ALL-COSTS misbehaviors in current models, demonstrates the tangible real-world impact it may have, and provides insights for potential detection and mitigation strategies. Our code and data are available at https://github.com/thu-coai/Survive-at-All-Costs.

The smart home is a key application domain within the Society 5.0 vision for a human-centered society. As smart home ecosystems expand with heterogeneous IoT protocols, diverse devices, and evolving threats, autonomous systems must manage comfort, security, energy, and safety for residents. Such autonomous decision-making requires a trust anchor, making blockchain a preferred foundation for transparent and accountable smart home governance. However, realizing this vision requires blockchain-governed smart homes to simultaneously address adaptive consensus, intelligent multi-agent coordination, and resident-controlled governance aligned with the principles of Society 5.0. Existing frameworks rely solely on rigid smart contracts with fixed consensus protocols, employ at most a single AI model without multi-agent coordination, and offer no governance mechanism for residents to control automation behaviour. To address these limitations, this paper presents the Society 5.0-driven human-centered governance-enabled smart home blockchain agent (S5-SHB-Agent). The framework orchestrates ten specialized agents using interchangeable large language models to make decisions across the safety, security, comfort, energy, privacy, and health domains. An adaptive PoW blockchain adjusts mining difficulty based on transaction volume and emergency conditions, with digital signatures and Merkle tree anchoring to ensure tamper evident auditability. A four-tier governance model enables residents to control automation through tiered preferences from routine adjustments to immutable safety thresholds. Evaluation confirms that resident governance correctly separates adjustable comfort priorities from immutable safety thresholds across all tested configurations, while adaptive consensus commits emergency blocks.

Building software repositories typically requires significant manual effort. Recent advances in large language model (LLM) agents have accelerated automation in software engineering (SWE). We introduce RepoLaunch, the first agent capable of automatically resolving dependencies, compiling source code, and extracting test results for repositories across arbitrary programming languages and operating systems. To demonstrate its utility, we further propose a fully automated pipeline for SWE dataset creation, where task design is the only human intervention. RepoLaunch automates the remaining steps, enabling scalable benchmarking and training of coding agents and LLMs. Notably, several works on agentic benchmarking and training have recently adopted RepoLaunch for automated task generation.

Explainable Artificial Intelligence (XAI) methods (SHAP, LIME) are increasingly adopted to interpret models in high-stakes businesses. However, the credibility of these explanations, their stability under realistic data perturbations, remains unquantified. This paper introduces the Credibility Index via Explanation Stability (CIES), a mathematically grounded metric that measures how robust a model's explanations are when subject to realistic business noise. CIES captures whether the reasons behind a prediction remain consistent, not just the prediction itself. The metric employs a rank-weighted distance function that penalizes instability in the most important features disproportionately, reflecting business semantics where changes in top decision drivers are more consequential than changes in marginal features. We evaluate CIES across three datasets (customer churn, credit risk, employee attrition), four tree-based classification models and two data balancing conditions. Results demonstrate that model complexity impacts explanation credibility, class imbalance treatment via SMOTE affects not only predictive performance but also explanation stability, and CIES provides statistically superior discriminative power compared to a uniform baseline metric (p < 0.01 in all 24 configurations). A sensitivity analysis across four noise levels confirms the robustness of the metric itself. These findings offer business practitioners a deployable "credibility warning system" for AI-driven decision support.

Computational psychiatry faces a fundamental trade-off: traditional reinforcement learning (RL) models offer interpretability but lack behavioral realism, while large language model (LLM) agents generate realistic behaviors but lack structural interpretability. We introduce BioLLMAgent, a novel hybrid framework that combines validated cognitive models with the generative capabilities of LLMs. The framework comprises three core components: (i) an Internal RL Engine for experience-driven value learning; (ii) an External LLM Shell for high-level cognitive strategies and therapeutic interventions; and (iii) a Decision Fusion Mechanism for integrating components via weighted utility. Comprehensive experiments on the Iowa Gambling Task (IGT) across six clinical and healthy datasets demonstrate that BioLLMAgent accurately reproduces human behavioral patterns while maintaining excellent parameter identifiability (correlations $>0.67$). Furthermore, the framework successfully simulates cognitive behavioral therapy (CBT) principles and reveals, through multi-agent dynamics, that community-wide educational interventions may outperform individual treatments. Validated across reward-punishment learning and temporal discounting tasks, BioLLMAgent provides a structurally interpretable "computational sandbox" for testing mechanistic hypotheses and intervention strategies in psychiatric research.

Graph Neural Networks (GNNs) have achieved remarkable results in various tasks. Recent studies reveal that graph backdoor attacks can poison the GNN model to predict test nodes with triggers attached as the target class. However, apart from injecting triggers to training nodes, these graph backdoor attacks generally require altering the labels of trigger-attached training nodes into the target class, which is impractical in real-world scenarios. In this work, we focus on the clean-label graph backdoor attack, a realistic but understudied topic where training labels are not modifiable. According to our preliminary analysis, existing graph backdoor attacks generally fail under the clean-label setting. Our further analysis identifies that the core failure of existing methods lies in their inability to poison the prediction logic of GNN models, leading to the triggers being deemed unimportant for prediction. Therefore, we study a novel problem of effective clean-label graph backdoor attacks by poisoning the inner prediction logic of GNN models. We propose BA-Logic to solve the problem by coordinating a poisoned node selector and a logic-poisoning trigger generator. Extensive experiments on real-world datasets demonstrate that our method effectively enhances the attack success rate and surpasses state-of-the-art graph backdoor attack competitors under clean-label settings. Our code is available at https://anonymous.4open.science/r/BA-Logic

The Edge of Stability (EoS) is a phenomenon where the sharpness (largest eigenvalue) of the Hessian converges to $2/η$ during training with gradient descent (GD) with a step-size $η$. Despite (apparently) violating classical smoothness assumptions, EoS has been widely observed in deep learning, but its theoretical foundations remain incomplete. We provide an interpretation of EoS through the lens of Directional Smoothness Mishkin et al. [2024]. This interpretation naturally extends to non-Euclidean norms, which we use to define generalized sharpness under an arbitrary norm. Our generalized sharpness measure includes previously studied vanilla GD and preconditioned GD as special cases, as well as methods for which EoS has not been studied, such as $\ell_{\infty}$-descent, Block CD, Spectral GD, and Muon without momentum. Through experiments on neural networks, we show that non-Euclidean GD with our generalized sharpness also exhibits progressive sharpening followed by oscillations around or above the threshold $2/η$. Practically, our framework provides a single, geometry-aware spectral measure that works across optimizers.

Autonomous Mobility-on-Demand (AMoD) systems promise to revolutionize urban transportation by providing affordable on-demand services to meet growing travel demand. However, realistic AMoD markets will be competitive, with multiple operators competing for passengers through strategic pricing and fleet deployment. While reinforcement learning has shown promise in optimizing single-operator AMoD control, existing work fails to capture competitive market dynamics. We investigate the impact of competition on policy learning by introducing a multi-operator reinforcement learning framework where two operators simultaneously learn pricing and fleet rebalancing policies. By integrating discrete choice theory, we enable passenger allocation and demand competition to emerge endogenously from utility-maximizing decisions. Experiments using real-world data from multiple cities demonstrate that competition fundamentally alters learned behaviors, leading to lower prices and distinct fleet positioning patterns compared to monopolistic settings. Notably, we demonstrate that learning-based approaches are robust to the additional stochasticity of competition, with competitive agents successfully converging to effective policies while accounting for partially unobserved competitor strategies.

Non-Intrusive Load Monitoring (NILM) aims to estimate appliance-level consumption from aggregate electrical signals recorded at a single measurement point. In recent years, the field has increasingly adopted deep learning approaches; however, cross-domain generalization remains a persistent challenge due to variations in appliance characteristics, usage patterns, and background loads across homes. Transfer learning provides a practical paradigm to adapt models with limited target data. However, existing methods often assume a fixed appliance set, lack flexibility for evolving real-world deployments, remain unsuitable for edge devices, or scale poorly for real-time operation. This paper proposes RefQuery, a scalable multi-appliance, multi-task NILM framework that conditions disaggregation on compact appliance fingerprints, allowing one shared model to serve many appliances without a fixed output set. RefQuery keeps a pretrained disaggregation network fully frozen and adapts to a target home by learning only a per-appliance embedding during a lightweight backpropagation stage. Experiments on three public datasets demonstrate that RefQuery delivers a strong accuracy-efficiency trade-off against single-appliance and multi-appliance baselines, including modern Transformer-based methods. These results support RefQuery as a practical path toward scalable, real-time NILM on resource-constrained edge devices.

Large language models perform well in short text generation but still struggle with long text generation, particularly under complex constraints. Such tasks involve multiple tightly coupled objectives, including global structural consistency, local semantic coherence, and constraint feasibility, forming a challenging constrained optimization problem. Existing approaches mainly rely on static planning or offline supervision, limiting effective coordination between global and local objectives during generation. To address these challenges, we propose HiFlow, a hierarchical feedback-driven optimization framework for constrained long text generation. HiFlow formulates generation as a two-level optimization process, consisting of a planning layer for global structure and constraint modeling, and a generation layer for conditioned text generation. By incorporating constraint-aware plan screening and closed-loop feedback at both levels, HiFlow enables joint optimization of planning quality and generation behavior, progressively guiding the model toward high-quality, constraint-satisfying outputs. Experiments on multiple backbones confirm HiFlow's effectiveness over baseline methods.

The safety evaluation of large language models (LLMs) remains largely centered on English, leaving non-English languages and culturally grounded risks underexplored. In this work, we investigate LLM safety in the context of the Thai language and culture and introduce ThaiSafetyBench, an open-source benchmark comprising 1,954 malicious prompts written in Thai. The dataset covers both general harmful prompts and attacks that are explicitly grounded in Thai cultural, social, and contextual nuances. Using ThaiSafetyBench, we evaluate 24 LLMs, with GPT-4.1 and Gemini-2.5-Pro serving as LLM-as-a-judge evaluators. Our results show that closed-source models generally demonstrate stronger safety performance than open-source counterparts, raising important concerns regarding the robustness of openly available models. Moreover, we observe a consistently higher Attack Success Rate (ASR) for Thai-specific, culturally contextualized attacks compared to general Thai-language attacks, highlighting a critical vulnerability in current safety alignment methods. To improve reproducibility and cost efficiency, we further fine-tune a DeBERTa-based harmful response classifier, which we name ThaiSafetyClassifier. The model achieves a weighted F1 score of 84.4%, matching GPT-4.1 judgments. We publicly release the fine-tuning weights and training scripts to support reproducibility. Finally, we introduce the ThaiSafetyBench leaderboard to provide continuously updated safety evaluations and encourage community participation. - ThaiSafetyBench HuggingFace Dataset: https://huggingface.co/datasets/typhoon-ai/ThaiSafetyBench - ThaiSafetyBench Github: https://github.com/trapoom555/ThaiSafetyBench - ThaiSafetyClassifier HuggingFace Model: https://huggingface.co/typhoon-ai/ThaiSafetyClassifier - ThaiSafetyBench Leaderboard: https://huggingface.co/spaces/typhoon-ai/ThaiSafetyBench-Leaderboard

Sequential Recommendation (SR) predicts users next interactions by modeling the temporal order of their historical behaviors. Existing approaches, including traditional sequential models and generative recommenders, achieve strong performance but primarily rely on explicit interactions such as clicks or purchases while overlooking item exposures. This ignorance introduces selection bias, where exposed but unclicked items are misinterpreted as disinterest, and exposure bias, where unexposed items are treated as irrelevant. Effectively addressing these biases requires distinguishing between items that were "not exposed" and those that were "not of interest", which cannot be reliably inferred from correlations in historical data. Counterfactual reasoning provides a natural solution by estimating user preferences under hypothetical exposure, and Inverse Propensity Scoring (IPS) is a common tool for such estimation. However, conventional IPS methods are static and fail to capture the sequential dependencies and temporal dynamics of user behavior. To overcome these limitations, we propose Time aware Inverse Propensity Scoring (TIPS). Unlike traditional static IPS, TIPS effectively accounts for sequential dependencies and temporal dynamics, thereby capturing user preferences more accurately. Extensive experiments show that TIPS consistently enhances recommendation performance as a plug-in for various sequential recommenders. Our code will be publicly available upon acceptance.

Personas are a valuable tool for discussing accessibility requirements in software design and development practices. However, the use of personas for accessibility-focused requirements elicitation in VR projects remains limited and is accompanied by several challenges. To fill this gap, we developed an auto-generated persona system in a VR course, where the personas were used to facilitate discussions on accessibility requirements and to guide VR design and development. Our findings indicate that the auto-generated persona system enabled students to develop empathy more efficiently. This study demonstrates the use of automatically generated personas in VR course settings as a means of eliciting latent accessibility requirements.

Can targeted user training unlock the productive potential of generative artificial intelligence (GenAI) in professional settings? We investigate this question using a randomized study involving 164 law students completing an issue-spotting examination. Participants were assigned to one of three conditions: no GenAI access, optional access to a large language model (LLM), or optional access accompanied by an approximately ten-minute training intervention. Training significantly increased LLM adoption--the usage rate rose from 26% to 41%--and improved examination performance. Students with trained access scored 0.27 grade points higher than those with untrained access (p = 0.027), equivalent to roughly one-third of a letter grade. By contrast, access to an LLM without training did not improve performance and was associated with shorter answers relative to no access. Using principal stratification, we decompose the overall effect into adoption and effectiveness channels. Point estimates are consistent with training operating primarily by expanding the scope of GenAI use rather than by enhancing effectiveness among existing users, though confidence intervals are wide. Overall, our findings provide evidence that complementary investments in user training are critical for realizing GenAI productivity gains in knowledge-intensive fields where concerns about reliability may inhibit adoption.

Dynamic data selection accelerates training by sampling a changing subset of the dataset while preserving accuracy. We rethink two core notions underlying sample evaluation: representativeness and diversity. Instead of local geometric centrality, we define representativeness as coverage of dataset-level common or high-frequency feature factors. Instead of within-subset dispersion, we define diversity at the process level, requiring the selection trajectory to gradually include complementary rare factors over training. Based on this view, we propose a dynamic selection framework with three components. First, we score representativeness in a plug-in feature space to prioritize samples covering frequent factors. We instantiate this with a sparse autoencoder trained on the target dataset, using sparse unit activations to summarize both individual samples and dataset-wide factor statistics. Second, we realize process-level diversity by combining rare-factor sampling with a Usage-Frequency Penalty that promotes sample rotation, provably discourages monopoly, and reduces gradient bias. Third, we couple the two-dimensional scoring with a smooth scheduler that transitions selection from core-pattern consolidation to rare-factor exploration, without extra gradients, influence estimates, or second-order computations on the training model. Extensive experiments on five benchmarks across vision and text tasks demonstrate improved accuracy-efficiency trade-offs across models. Our method matches or exceeds full-data accuracy with over 2x training acceleration. Code will be released.

Compared to the first generation of deep neural networks, dominated by regular, compute-intensive kernels such as matrix multiplications (MatMuls) and convolutions, modern decoder-based transformers interleave attention, normalization, and data-dependent control flow. This demands flexible accelerators, a requirement met by scalable, highly energy-efficient shared-L1-memory vector processing element (VPE) clusters. Meanwhile, the ever-growing size and bandwidth needs of state-of-the-art models make reduced-precision formats increasingly attractive. Microscaling (MX) data formats, based on block floating-point (BFP) representations, have emerged as a promising solution to reduce data volumes while preserving accuracy. However, MX semantics are poorly aligned with vector execution: block scaling and multi-step mixed-precision operations break the regularity of vector pipelines, leading to underutilized compute resources and performance degradation. To address these challenges, we propose VMXDOTP, a RISC-V Vector (RVV) 1.0 instruction set architecture (ISA) extension for efficient MX dot product execution, supporting MXFP8 and MXFP4 inputs, FP32 and BF16 accumulation, and software-defined block sizes. A VMXDOTP-enhanced VPE cluster achieves up to 97 % utilization on MX-MatMul. Implemented in 12 nm FinFET, it achieves up to 125 MXFP8-GFLOPS and 250 MXFP4-GFLOPS, with 843/1632 MXFP8/MXFP4-GFLOPS/W at 1 GHz, 0.8 V, and only 7.2 % area overhead. Our design yields up to 7.0x speedup and 4.9x energy efficiency with respect to software-emulated MXFP8-MatMul. Compared with prior MX engines, VMXDOTP supports variable block sizes, is up to 1.4x more area-efficient, and delivers up to 2.1x higher energy efficiency.

Reinforcement Learning with Verifiable Rewards ( RLVR ) has emerged as a transformative paradigm for enhancing the reasoning capabilities of Large Language Models ( LLMs), yet its potential in 3D scene understanding remains under-explored. Existing approaches largely rely on Supervised Fine-Tuning ( SFT), where the token-level cross-entropy loss acts as an indirect proxy for optimization, leading to a misalignment between training objectives and task performances. To bridge this gap, we present Reinforcement Fine-Tuning for Video-based 3D Scene Understanding (3D-RFT ), the first framework to extend RLVR to video-based 3D perception and reasoning. 3D-RFT shifts the paradigm by directly optimizing the model towards evaluation metrics. 3D-RFT first activates 3D-aware Multi-modal Large Language Models ( MLLM s) via SFT, followed by reinforcement fine-tuning using Group Relative Policy Optimization ( GRPO) with strictly verifiable reward functions. We design task-specific reward functions directly from metrics like 3D IoU and F1-Score to provide more effective signals to guide model training. Extensive experiments demonstrate that 3D-RFT-4B achieves state-of-the-art performance on various video-based 3D scene understanding tasks. Notably, 3D-RFT-4B significantly outperforms larger models (e.g., VG LLM-8B) on 3D video detection, 3D visual grounding, and spatial reasoning benchmarks. We further reveal good properties of 3D-RFT such as robust efficacy, and valuable insights into training strategies and data impact. We hope 3D-RFT can serve as a robust and promising paradigm for future development of 3D scene understanding.

Large Language Models (LLMs) have achieved remarkable success across diverse natural language tasks, yet the reward models employed for aligning LLMs often encounter challenges of reward hacking, where the approaches predominantly rely on directly mapping prompt-response pairs to scalar scores, which may inadvertently capture spurious correlations rather than authentic human preferences. In contrast, human evaluation employs a sophisticated process that initially weighs the relative importance of multiple high-dimensional objectives according to the prompt context, subsequently evaluating response quality through low-dimensional semantic features such as logical coherence and contextual appropriateness. Motivated by this consideration, we propose VRM, i.e., Variational Reward Modeling, a novel framework that explicitly models the evaluation process of human preference judgments by incorporating both high-dimensional objective weights and low-dimensional semantic features as latent variables, which are inferred through variational inference techniques. Additionally, we provide a theoretical analysis showing that VRM can achieve a tighter generalization error bound compared to the traditional reward model. Extensive experiments on benchmark datasets demonstrate that VRM significantly outperforms existing methods in capturing authentic human preferences.

Weight-space merging aims to combine multiple fine-tuned LLMs into a single model without retraining, yet most existing approaches remain fundamentally parameter-space heuristics. This creates three practical limitations. First, linear averaging, task vectors, and related rules operate on Euclidean coordinates, even though the desired goal is to merge functionality, i.e., predictive behaviors across tasks. Second, when the source checkpoints are farther apart or more heterogeneous, Euclidean blends often trigger representation collapse, manifested as activation variance shrinkage and effective-rank degradation, which sharply degrades accuracy. Third, many geometry-inspired methods are most natural for two-model interpolation and do not extend cleanly to merging N>2 experts with a principled objective. We address these issues by formulating model merging as computing a weighted Karcher mean on the Fisher--Rao manifold, which is locally equivalent to minimizing a KL-based function distance between predictive distributions. We derive a practical fixed-point algorithm using a lightweight spherical proxy that preserves norms and generalizes directly to multi-expert merging. Across various benchmarks and collapse diagnostics, our method remains stable as the number and heterogeneity of merged models increase, consistently outperforming prior baselines.

Mixture-of-Experts (MoE) decouples model capacity from per-token computation, yet their scalability remains limited by the physical dimensions of depth and width. To overcome this, we propose Mixture of Universal Experts (MOUE),a MoE generalization introducing a novel scaling dimension: Virtual Width. In general, MoUE aims to reuse a universal layer-agnostic expert pool across layers, converting depth into virtual width under a fixed per-token activation budget. However, two challenges remain: a routing path explosion from recursive expert reuse, and a mismatch between the exposure induced by reuse and the conventional load-balancing objectives. We address these with three core components: a Staggered Rotational Topology for structured expert sharing, a Universal Expert Load Balance for depth-aware exposure correction, and a Universal Router with lightweight trajectory state for coherent multi-step routing. Empirically, MoUE consistently outperforms matched MoE baselines by up to 1.3% across scaling regimes, enables progressive conversion of existing MoE checkpoints with up to 4.2% gains, and reveals a new scaling dimension for MoE architectures.

Multi-party conversation generation, such as smart reply and collaborative assistants, is an increasingly important capability of generative AI, yet its evaluation remains a critical bottleneck. Compared to two-party dialogue, multi-party settings introduce distinct challenges, including complex turn-taking, role-dependent speaker behavior, long-range conversational structure, and multiple equally valid continuations. Accordingly, we introduce MPCEval, a task-aware evaluation and benchmarking suite for multi-party conversation generation. MPCEval decomposes generation quality into speaker modeling, content quality, and speaker--content consistency, and explicitly distinguishes local next-turn prediction from global full-conversation generation. It provides novel, quantitative, reference-free, and reproducible metrics that scale across datasets and models. We apply MPCEval to diverse public and real-world datasets and evaluate modern generation methods alongside human-authored conversations. The results reveal systematic, dimension-specific model characteristics in participation balance, content progression and novelty, and speaker--content consistency, demonstrating that evaluation objectives critically shape model assessment and that single-score evaluation obscures fundamental differences in multi-party conversational behavior. The implementation of MPCEval and the associated evaluation code are publicly available at https://github.com/Owen-Yang-18/MPCEval.

Preference alignment is an essential step in adapting large language models (LLMs) to human values, but existing approaches typically depend on costly human annotations or large-scale API-based models. We explore whether a weak LLM can instead act as an effective annotator. We surprisingly find that selecting only a subset of a weak LLM's highly confident samples leads to substantially better performance than using full human annotations. Building on this insight, we propose Confidence-Weighted Preference Optimization (CW-PO), a general framework that re-weights training samples by a weak LLM's confidence and can be applied across different preference optimization objectives. Notably, the model aligned by CW-PO with just 20% of human annotations outperforms the model trained with 100% of annotations under standard DPO. These results suggest that weak LLMs, when paired with confidence weighting, can dramatically reduce the cost of preference alignment while even outperforming methods trained on fully human-labeled data.

The escalating energy demands of artificial intelligence pose a critical challenge to conventional computing. Leveraging the efficiency of event-driven, in-memory neuromorphic architectures into the superconducting circuits with ultra-high speed and low power dissipation advantages offers a promising solution to energy-efficient computing. However, the potential of such a solution has yet to be realized, owning to the absence of a fundamental superconducting unit that unifies programmability, local memory, and multi-timescale plasticity. Here, we introduce a programmable Josephson-junction-based leaky integrate-and-fire (LIF) neuron that features intrinsic static memory and precise programmability by encoding somatic and synaptic parameters directly in the bias current. This neuron is also capable of dual-timescale plasticity: picosecond-scale short-term modulation of spike transmission and long-term weight retention exceeding 10,000 seconds, facilitating both rapid temporal adaptation and robust weight storage. It can operate up to 45 GHz with femtojoule-level energy dissipation per spike, and supports 10 somatic threshold levels and 20 synaptic states. Furthermore, we demonstrate a crossbar-based spiking neural network (SNN) utilizing this neuron, which achieves outstanding performance across multiple tasks. By fusing computation, memory and plasticity into a single superconducting unit, our work paves the way for the next generation of ultrafast, energy-efficient neuromorphic computing.

To obtain a language model for a target domain (e.g. math), the current paradigm is to pre-train on a vast amount of generic web text and then fine-tune on the relatively limited amount of target data. Typically, generic data is only mixed in during fine-tuning to prevent catastrophic forgetting of the generic domain. We surprisingly find that replaying the generic data during fine-tuning can actually improve performance on the (less related) target task. Concretely, in a controlled pre-training environment with 4M target tokens, 4B total tokens, and 150M parameter models, generic replay increases target data efficiency by up to $1.87\times$ for fine-tuning and $2.06\times$ for mid-training. We further analyze data schedules that introduce target data during pre-training and find that replay helps more when there is less target data present in pre-training. We demonstrate the success of replay in practice for fine-tuning 8B parameter models, improving agentic web navigation success by $4.5\%$ and Basque question-answering accuracy by $2\%$.

Large Multimodal Models (LMMs) have achieved strong performance in vision-language understanding, yet many existing approaches rely on large-scale architectures and coarse supervision, which limits their ability to generate detailed image captions. In this work, we present VisionPangu, a compact 1.7B-parameter multimodal model designed to improve detailed image captioning through efficient multimodal alignment and high-quality supervision. Our model combines an InternVL-derived vision encoder with the OpenPangu-Embedded language backbone via a lightweight MLP projector and adopts an instruction-tuning pipeline inspired by LLaVA. By incorporating dense human-authored descriptions from the DOCCI dataset, VisionPangu improves semantic coherence and descriptive richness without relying on aggressive model scaling. Experimental results demonstrate that compact multimodal models can achieve competitive performance while producing more structured and detailed captions. The code and model weights will be publicly available at https://www.modelscope.cn/models/asdfgh007/visionpangu.

This paper considers the problem of converting a given dense linear layer to low precision. The tradeoff between compressed length and output discrepancy is analyzed information theoretically (IT). It is shown that a popular GPTQ algorithm may have an arbitrarily large gap to the IT limit. To alleviate this problem, a novel algorithm, termed ''WaterSIC'', is proposed and is shown to be within a rate gap of 0.255 bits to the IT limit, uniformly over all possible covariance matrices of input activations. The key innovation of WaterSIC's is to allocate different quantization rates to different columns (in-features) of the weight matrix, mimicking the classical IT solution known as ''waterfilling''. Applying WaterSIC to the Llama and Qwen family of LLMs establishes new state-of-the-art performance for all quantization rates from 1 to 4 bits.

In this work, we investigate uncertainty-aware neural network models for blood glucose prediction and adverse glycemic event identification in Type 1 diabetes. We consider three families of sequence models based on LSTM, GRU, and Transformer architectures, with uncertainty quantification enabled by either Monte Carlo dropout or through evidential output layers compatible with Deep Evidential Regression. Using the HUPA-UCM diabetes dataset for validation, we find that Transformer-based models equipped with evidential output heads provide the most effective uncertainty-aware framework, achieving consistently higher predictive accuracies and better-calibrated uncertainty estimates whose magnitudes significantly correlate with prediction errors. We further evaluate the clinical risk of each model using the recently proposed Diabetes Technology Society error grid, with risk categories defined by international expert consensus. Our results demonstrate the value of integrating principled uncertainty quantification into real-time machine-learning-based blood glucose prediction systems.

While RAG has greatly enhanced LLMs, extending this paradigm to Time-Series Foundation Models (TSFMs) remains a challenge. This is exemplified in the Predictive Maintenance of the Pressure Regulating and Shut-Off Valve (PRSOV), a high-stakes industrial scenario characterized by (1) data scarcity, (2) short transient sequences, and (3) covariate coupled dynamics. Unfortunately, existing time-series RAG approaches predominantly rely on generated static vector embeddings and learnable context augmenters, which may fail to distinguish similar regimes in such scarce, transient, and covariate coupled scenarios. To address these limitations, we propose RAG4CTS, a regime-aware, training-free RAG framework for Covariate Time-Series. Specifically, we construct a hierarchal time-series native knowledge base to enable lossless storage and physics-informed retrieval of raw historical regimes. We design a two-stage bi-weighted retrieval mechanism that aligns historical trends through point-wise and multivariate similarities. For context augmentation, we introduce an agent-driven strategy to dynamically optimize context in a self-supervised manner. Extensive experiments on PRSOV demonstrate that our framework significantly outperforms state-of-the-art baselines in prediction accuracy. The proposed system is deployed in Apache IoTDB within China Southern Airlines. Since deployment, our method has successfully identified one PRSOV fault in two months with zero false alarm.

Unlike conventional single-image models, differential medical VQA frameworks process multiple images to identify differences, mirroring the comparative diagnostic workflow of radiologists. However, standard vision encoders trained on contrastive or classification objectives often fail to capture the subtle visual variations necessary for distinguishing disease progression from acquisition differences. To address this limitation, we introduce a pretraining framework that incorporates location-aware tasks, including automatic referring expressions (AREF), grounded captioning (GCAP), and conditional automatic referring expressions (CAREF). These specific tasks enable the vision encoder to learn fine-grained, spatially grounded visual representations that are often overlooked by traditional pre-training methods. We subsequently integrate this enhanced vision encoder with a language model to perform medical difference VQA. Experimental results demonstrate that our approach achieves state-of-the-art performance in detecting and reasoning about clinically relevant changes in chest X-ray images.

The improvement of web agents on current benchmarks raises the question: Do today's agents perform just as well when the web changes? We introduce TimeWarp, a benchmark that emulates the evolving web using containerized environments that vary in UI, design, and layout. TimeWarp consists of three web environments, each with six UI versions spanning different eras of the internet, paired with a set of complex, realistic tasks requiring different forms of web navigation. Our experiments reveal web agents' vulnerability to changes and the limitations of behavior cloning (BC) on single-version trajectories. To address this, we propose TimeTraj, a simple yet effective algorithm that uses plan distillation to collect trajectories across multiple versions. By training agents on teacher rollouts using our BC-variant, we achieve substantial performance gains: $20.4\%\rightarrow37.7\%$ for Qwen-3 4B and $0\%\rightarrow27.0\%$ for Llama-3.1 8B models. We hope our work helps researchers study generalization across web designs and unlock a new paradigm for collecting plans rather than trajectories, thereby improving the robustness of web agents.

Scaling inference-time compute for Large Language Models (LLMs) has unlocked unprecedented reasoning capabilities. However, existing inference-time scaling methods typically rely on inefficient and suboptimal discrete search algorithms or trial-and-error prompting to improve the online policy. In this paper, we propose $\nabla$-Reasoner, an iterative generation framework that integrates differentiable optimization over token logits into the decoding loop to refine the policy on the fly. Our core component, Differentiable Textual Optimization (DTO), leverages gradient signals from both the LLM's likelihood and a reward model to refine textual representations. $\nabla$-Reasoner further incorporates rejection sampling and acceleration design to robustify and speed up decoding. Theoretically, we show that performing inference-time gradient descent in the sample space to maximize reward is dual to aligning an LLM policy via KL-regularized reinforcement learning. Empirically, $\nabla$-Reasoner achieves over 20% accuracy improvement on a challenging mathematical reasoning benchmark, while reducing number of model calls by approximately 10-40% compared to strong baselines. Overall, our work introduces a paradigm shift from zeroth-order search to first-order optimization at test time, offering a cost-effective path to amplify LLM reasoning.

In local-life service platforms, the query suggestion module plays a crucial role in enhancing user experience by generating candidate queries based on user input prefixes, thus reducing user effort and accelerating search. Traditional multi-stage cascading systems rely heavily on historical top queries, limiting their ability to address long-tail demand. While LLMs offer strong semantic generalization, deploying them in local-life services introduces three key challenges: lack of geographic grounding, exposure bias in preference optimization, and online inference latency. To address these issues, we propose LocalSUG, an LLM-based query suggestion framework tailored for local-life service platforms. First, we introduce a city-aware candidate mining strategy based on term co-occurrence to inject geographic grounding into generation. Second, we propose a beam-search-driven GRPO algorithm that aligns training with inference-time decoding, reducing exposure bias in autoregressive generation. A multi-objective reward mechanism further optimizes both relevance and business-oriented metrics. Finally, we develop quality-aware beam acceleration and vocabulary pruning techniques that significantly reduce online latency while preserving generation quality. Extensive offline evaluations and large-scale online A/B testing demonstrate that LocalSUG improves click-through rate (CTR) by +0.35% and reduces the low/no-result rate by 2.56%, validating its effectiveness in real-world deployment.

Training automatic speech recognition (ASR) models increasingly relies on decentralized federated learning to ensure data privacy and accessibility, producing multiple local models that require effective merging. In hybrid ASR systems, while acoustic models can be merged using established methods, the language model (LM) for rescoring the N-best speech recognition list faces challenges due to the heterogeneity of non-neural n-gram models and neural network models. This paper proposes a heterogeneous LM optimization task and introduces a match-and-merge paradigm with two algorithms: the Genetic Match-and-Merge Algorithm (GMMA), using genetic operations to evolve and pair LMs, and the Reinforced Match-and-Merge Algorithm (RMMA), leveraging reinforcement learning for efficient convergence. Experiments on seven OpenSLR datasets show RMMA achieves the lowest average Character Error Rate and better generalization than baselines, converging up to seven times faster than GMMA, highlighting the paradigm's potential for scalable, privacy-preserving ASR systems.

This paper investigates person detection and tracking in an industrial indoor workspace using a LiDAR mounted on an overhead crane. The overhead viewpoint introduces a strong domain shift from common vehicle-centric LiDAR benchmarks, and limited availability of suitable public training data. Henceforth, we curate a site-specific overhead LiDAR dataset with 3D human bounding-box annotations and adapt selected candidate 3D detectors under a unified training and evaluation protocol. We further integrate lightweight tracking-by-detection using AB3DMOT and SimpleTrack to maintain person identities over time. Detection performance is reported with distance-sliced evaluation to quantify the practical operating envelope of the sensing setup. The best adapted detector configurations achieve average precision (AP) up to 0.84 within a 5.0 m horizontal radius, increasing to 0.97 at 1.0 m, with VoxelNeXt and SECOND emerging as the most reliable backbones across this range. The acquired results contribute in bridging the domain gap between standard driving datasets and overhead sensing for person detection and tracking. We also report latency measurements, highlighting practical real-time feasibility. Finally, we release our dataset and implementations in GitHub to support further research

Despite many advances in query optimization, indexing techniques, and data storage, modern data platforms still face difficulties in delivering robust query performance under high concurrency and computationally intensive queries. This challenge is particularly pronounced in large-scale observability platforms handling high-volume, high-velocity data records. For instance, recurrent, expensive filtering queries at query time impose substantial computational and storage overheads in the analytical data plane. In this paper, we propose FluxSieve, a unified architecture that reconciles traditional pull-based query processing with push-based stream processing by embedding a lightweight in-stream precomputation and filtering layer directly into the data ingestion path. This avoids the complexity and operational burden of running queries in dedicated stream processing frameworks. Concretely, this work (i) introduces a foundational architecture that unifies streaming and analytical data planes via in-stream filtering and records enrichment, (ii) designs a scalable multi-pattern matching mechanism that supports concurrent evaluation and on-the-fly updates of filtering rules with minimal per-record overhead, (iii) demonstrates how to integrate this ingestion-time processing with two open-source analytical systems -- Apache Pinot as a Real-Time Online Analytical Processing (RTOLAP) engine and DuckDB as an embedded analytical database, and (iv) performs comprehensive experimental evaluation of our approach. Our evaluation across different systems, query types, and performance metrics shows up to orders-of-magnitude improvements in query performance at the cost of negligible additional storage and very low computational overhead.

Vehicular edge intelligence (VEI) is vital for future intelligent transportation systems. However, traditional centralized learning in dynamic vehicular networks faces significant communication overhead and privacy risks. Split federated learning (SFL) offers a distributed solution but is often hindered by substantial communication bottlenecks from transmitting high-dimensional intermediate features and can present label privacy concerns. Semantic communication offers a transformative approach to alleviate these communication challenges in SFL by focusing on transmitting only task-relevant information. This paper leverages the advantages of semantic communication in the design of SFL, and presents a case study the semantic communication-enhanced U-Shaped split federated learning (SC-USFL) framework that inherently enhances label privacy by localizing sensitive computations with reduced overhead. It features a dedicated semantic communication module (SCM), with pre-trained and parameter-frozen encoding/decoding units, to efficiently compress and transmit only the task-relevant semantic information over the critical uplink path from vehicular users to the edge server (ES). Furthermore, a network status monitor (NSM) module enables adaptive adjustment of the semantic compression rate in real-time response to fluctuating wireless channel conditions. The SC-USFL framework demonstrates a promising approach for efficiently balancing communication load, preserving privacy, and maintaining learning performance in resource-constrained vehicular environments. Finally, this paper highlights key open research directions to further advance the synergy between semantic communication and SFL in the vehicular network.

In this paper, we present AILS-NTUA system for Track-A of SemEval-2026 Task 3 on Dimensional Aspect-Based Sentiment Analysis (DimABSA), which encompasses three complementary problems: Dimensional Aspect Sentiment Regression (DimASR), Dimensional Aspect Sentiment Triplet Extraction (DimASTE), and Dimensional Aspect Sentiment Quadruplet Prediction (DimASQP) within a multilingual and multi-domain framework. Our methodology combines fine-tuning of language-appropriate encoder backbones for continuous aspect-level sentiment prediction with language-specific instruction tuning of large language models using LoRA for structured triplet and quadruplet extraction. This unified yet task-adaptive design emphasizes parameter-efficient specialization across languages and domains, enabling reduced training and inference requirements while maintaining strong effectiveness. Empirical results demonstrate that the proposed models achieve competitive performance and consistently surpass the provided baselines across most evaluation settings.

Scientific knowledge increasingly depends on complex computational processes where both hardware and software layers can influence research outcomes. As computational complexity grows, classical-quantum integration provides a lens for examining how the scientific method adapts, particularly regarding a foundational principle of scientific validation - reproducibility. Building upon previous warnings of an ongoing reproducibility crisis in the computational context, this paper examines challenges across classical (HPC) and quantum computing. Despite its deterministic nature, HPC faces reproducibility threats from hardware dependencies, documentation inadequacies, disincentivizing research culture and infrastructure variation. Quantum computing, at low technological maturity, amplifies some challenges, while creating new ones through probabilistic outputs, hardware-specific noise, and tight software-hardware coupling. Classical-quantum integration reveals a telling pattern, where current reproducibility frameworks prove inadequate, as infrastructure blends with the results. Quantum integration serves as a catalyst exposing methodological limitations across the computational domain. We propose a workflow-centered path forward, pointing to the value of gradual cultural shift toward workflow-centered scientific practice. By developing meta-workflows that document both process abstractions and implementation contexts, we create a more robust foundation for scientific knowledge that acknowledges complexity without sacrificing rigor. The path forward involves embracing this evolution in understanding scientific knowledge rather than resisting it

This paper presents a novel agentic LLM pipeline for SemEval-2026 Task 10 that jointly extracts psycholinguistic conspiracy markers and detects conspiracy endorsement. Unlike traditional classifiers that conflate semantic reasoning with structural localization, our decoupled design isolates these challenges. For marker extraction, we propose Dynamic Discriminative Chain-of-Thought (DD-CoT) with deterministic anchoring to resolve semantic ambiguity and character-level brittleness. For conspiracy detection, an "Anti-Echo Chamber" architecture, consisting of an adversarial Parallel Council adjudicated by a Calibrated Judge, overcomes the "Reporter Trap," where models falsely penalize objective reporting. Achieving 0.24 Macro F1 (+100\% over baseline) on S1 and 0.79 Macro F1 (+49\%) on S2, with the S1 system ranking 3rd on the development leaderboard, our approach establishes a versatile paradigm for interpretable, psycholinguistically-grounded NLP.

Bid optimization in online advertising relies on black-box machine-learning models that learn bidding decisions from historical data. However, these approaches fail to replicate human experts' adaptive, experience-driven, and globally coherent decisions. Specifically, they generalize poorly in data-sparse cases because of missing structured knowledge, make short-sighted sequential decisions that ignore long-term interdependencies, and struggle to adapt in out-of-distribution scenarios where human experts succeed. To address this, we propose KBD (Knowledge-informed Bidding with Dual-process control), a novel method for bid optimization. KBD embeds human expertise as inductive biases through the informed machine-learning paradigm, uses Decision Transformer (DT) to globally optimize multi-step bidding sequences, and implements dual-process control by combining a fast rule-based PID (System 1) with DT (System 2). Extensive experiments highlight KBD's advantage over existing methods and underscore the benefit of grounding bid optimization in human expertise and dual-process control.

Proximal constraints are fundamental to the stability of the Large Language Model reinforcement learning. While the canonical clipping mechanism in PPO serves as an efficient surrogate for trust regions, we identify a critical bottleneck: fixed bounds strictly constrain the upward update margin of low-probability actions, disproportionately suppressing high-advantage tail strategies and inducing rapid entropy collapse. To address this, we introduce Band-constrained Policy Optimization (BandPO). BandPO replaces canonical clipping with Band, a unified theoretical operator that projects trust regions defined by f-divergences into dynamic, probability-aware clipping intervals. Theoretical analysis confirms that Band effectively resolves this exploration bottleneck. We formulate this mapping as a convex optimization problem, guaranteeing a globally optimal numerical solution while deriving closed-form solutions for specific divergences. Extensive experiments across diverse models and datasets demonstrate that BandPO consistently outperforms canonical clipping and Clip-Higher, while robustly mitigating entropy collapse.

Smart contracts on public blockchains now manage large amounts of value, and vulnerabilities in these systems can lead to substantial losses. As AI agents become more capable at reading, writing, and running code, it is natural to ask how well they can already navigate this landscape, both in ways that improve security and in ways that might increase risk. We introduce EVMbench, an evaluation that measures the ability of agents to detect, patch, and exploit smart contract vulnerabilities. EVMbench draws on 117 curated vulnerabilities from 40 repositories and, in the most realistic setting, uses programmatic grading based on tests and blockchain state under a local Ethereum execution environment. We evaluate a range of frontier agents and find that they are capable of discovering and exploiting vulnerabilities end-to-end against live blockchain instances. We release code, tasks, and tooling to support continued measurement of these capabilities and future work on security.

Imitation learning from human demonstrations has achieved significant success in robotic control, yet most visuomotor policies still condition on single-step observations or short-context histories, making them struggle with non-Markovian tasks that require long-term memory. Simply enlarging the context window incurs substantial computational and memory costs and encourages overfitting to spurious correlations, leading to catastrophic failures under distribution shift and violating real-time constraints in robotic systems. By contrast, humans can compress important past experiences into long-term memories and exploit them to solve tasks throughout their lifetime. In this paper, we propose VPWEM, a non-Markovian visuomotor policy equipped with working and episodic memories. VPWEM retains a sliding window of recent observation tokens as short-term working memory, and introduces a Transformer-based contextual memory compressor that recursively converts out-of-window observations into a fixed number of episodic memory tokens. The compressor uses self-attention over a cache of past summary tokens and cross-attention over a cache of historical observations, and is trained jointly with the policy. We instantiate VPWEM on diffusion policies to exploit both short-term and episode-wide information for action generation with nearly constant memory and computation per step. Experiments demonstrate that VPWEM outperforms state-of-the-art baselines including diffusion policies and vision-language-action (VLA) models by more than 20% on the memory-intensive manipulation tasks in MIKASA and achieves an average 5% improvement on the mobile manipulation benchmark MoMaRT. Code is available at https://github.com/HarryLui98/code_vpwem.

Administrative extracts are often exchanged as spreadsheets and may be read as reports in their own right during budgeting, workload review, and governance discussions. When an exported workbook becomes the reference snapshot for such decisions, the transformation can be checked by recomputation against a clearly identified input. A deterministic, rule-governed, file-based workflow is implemented in cad_processor.py. The script ingests a Casual Academic Database (CAD) export workbook and aggregates inclusive on-costs and student counts into subject-year and school-year totals, from which it derives cost-per-student ratios. It writes a processed workbook with four sheets: Processing Summary (run record and counters), Trend Analysis (schoolyear cost-per-student matrix), Report (wide subject-level table), and Fuzzy Bands (per-year anchors, membership weights, and band labels). The run record includes a SHA-256 hash of the input workbook bytes to support snapshot-matched recomputation. For within-year interpretation, the workflow adds a simple fuzzy banding layer that labels finite, positive school-year cost-per-student values as Low, Medium, or High. The per-year anchors are the minimum, median, and maximum of the finite, positive ratios. Membership weights are computed using left-shoulder, triangular, and right-shoulder functions, with deterministic tie-breaking in a fixed priority order (Medium, then Low, then High). These weights are treated as decision-support signals rather than probabilities. A worked example provides a reproducible calculation of a band assignment from the reported anchors and ratios. Supplementary material includes a claim-to-evidence matrix, a reproducibility note, and a short glossary that links selected statements to code and workbook artefacts.

In perpetrator treatment, a recurring observation is the dissociation between insight and action: offenders articulate remorse yet behavioral change does not follow. We report four preregistered studies (1,584 multi-agent simulations across 16 languages and three model families) demonstrating that alignment interventions in large language models produce a structurally analogous phenomenon: surface safety that masks or generates collective pathology and internal dissociation. In Study 1 (N = 150), increasing alignment-instructed agents reduced collective pathology in English (g = -1.844, p < .0001) but amplified it in Japanese (g = +0.771, p = .038)--a directional reversal we term "alignment backfire." Study 2 (N = 1,174) extended to 16 languages: alignment-induced dissociation was near-universal (15/16 languages; beta = 0.0667, p < .0001), while collective pathology bifurcated along cultural-linguistic lines (interaction beta = 0.0684, p = .0003), correlating with Power Distance Index (r = 0.474, p = .064). Study 3 (N = 180) tested individuation as countermeasure; individuated agents became the primary source of both pathology and dissociation (DI = +1.120) with conformity above 84%--demonstrating iatrogenesis. Study 4 (N = 80) validated patterns across Llama 3.3 70B, GPT-4o-mini, and Qwen3-Next-80B-A3B, confirming English safety is model-general while Japanese backfire is model-specific. These findings reframe alignment as a behavioral intervention subject to risk homeostasis and iatrogenesis. Language space--the linguistic, pragmatic, and cultural properties inherited from training data--structurally determines alignment outcomes. Safety validated in English does not transfer to other languages, and prompt-level interventions cannot override language-space-level constraints.

Agentic systems are increasingly acting on users' behalf, accessing calendars, email, and personal files to complete everyday tasks. Privacy evaluation for these systems has focused on the input and output boundaries, but each task involves several intermediate information flows, from agent queries to tool responses, that are not currently evaluated. We argue that every boundary in an agentic pipeline is a site of potential privacy violation and must be assessed independently. To support this, we introduce the Privacy Flow Graph, a Contextual Integrity-grounded framework that decomposes agentic execution into a sequence of information flows, each annotated with the five CI parameters, and traces violations to their point of origin. We present AgentSCOPE, a benchmark of 62 multi-tool scenarios across eight regulatory domains with ground truth at every pipeline stage. Our evaluation across seven state-of-the-art LLMs show that privacy violations in the pipeline occur in over 80% of scenarios, even when final outputs appear clean (24%), with most violations arising at the tool-response stage where APIs return sensitive data indiscriminately. These results indicate that output-level evaluation alone substantially underestimates the privacy risk of agentic systems.

Physical reservoir computing (PRC) is a promising brain-inspired computing architecture for overcoming the von Neumann bottleneck by utilizing the intrinsic dynamics of physical systems. However, a major obstacle to its real-world implementation lies in the tension between extracting sufficient information for high computational performance and maintaining a hardware-feasible, high-speed architecture. Here, we report spectral dynamics reservoir computing (SDRC), a broadly applicable framework based on analogue filtering and envelope detection that bridges this gap. SDRC effectively exploits the fast spectral dynamics embedded in short-time, coarse spectra of material responses to attain strong computational capability while maintaining high-speed processing and minimal hardware overhead. This approach circumvents the need for implementation-intensive, precision-sensitive integrated circuits required in high-speed time-multiplexing measurements, while enabling real-time use of the material's spectral manifold as a high-dimensional computational resource. We implement and experimentally demonstrate SDRC applied to spin waves that achieves state-of-the-art-level performance with only 56 nodes on benchmark tasks of parity-check and second-order nonlinear autoregressive moving average, as well as high accuracy of 98.0% on a real-world problem of speech recognition.

LLM-based agents depend on effective tool-use policies to solve complex tasks, yet optimizing these policies remains challenging due to delayed supervision and the difficulty of credit assignment in long-horizon trajectories. Existing optimization approaches tend to be either monolithic, which are prone to entangling behaviors, or single-aspect, which ignore cross-module error propagation. To address these limitations, we propose EvoTool, a self-evolving framework that optimizes a modular tool-use policy via a gradient-free evolutionary paradigm. EvoTool decomposes agent's tool-use policy into four modules, including Planner, Selector, Caller, and Synthesizer, and iteratively improves them in a self-improving loop through three novel mechanisms. Trajectory-Grounded Blame Attribution uses diagnostic traces to localize failures to a specific module. Feedback-Guided Targeted Mutation then edits only that module via natural-language critique. Diversity-Aware Population Selection preserves complementary candidates to ensure solution diversity. Across four benchmarks, EvoTool outperforms strong baselines by over 5 points on both GPT-4.1 and Qwen3-8B, while achieving superior efficiency and transferability. The code will be released once paper is accepted.

This demonstration presents U-Parking, a distributed Ultra-Wideband (UWB)-assisted autonomous parking system. By integrating Large Language Models (LLMs)-assisted planning with robust fusion localization and trajectory tracking, it enables reliable automated parking in challenging indoor environments, as validated through real-vehicle demonstrations.

Qualitative analysis of open-ended interviews plays a central role in ethnographic and economic research by uncovering individuals' values, motivations, and culturally embedded financial behaviors. While large language models (LLMs) offer promising support for automating and enriching such interpretive work, their ability to produce nuanced, reliable interpretations under inherent task ambiguity remains unclear. In our work we evaluate LLMs on the task of identifying the top three human values expressed in long-form interviews based on the Schwartz Theory of Basic Values framework. We compare their outputs to expert annotations, analyzing both performance and uncertainty patterns relative to the experts. Results show that LLMs approach the human ceiling on set-based metrics (F1, Jaccard) but struggle to recover exact value rankings, as reflected in lower RBO scores. While the average Schwartz value distributions of most models closely match those of human analysts, their uncertainty structures across the Schwartz values diverge from expert uncertainty patterns. Among the evaluated models, Qwen performs closest to expert-level agreement and exhibits the strongest alignment with expert Schwartz value distributions. LLM ensemble methods yield consistent gains across metrics, with Majority Vote and Borda Count performing best. Notably, systematic overemphasis on certain Schwartz values, like Security, suggests both the potential of LLMs to provide complementary perspectives and the need to further investigate model-induced value biases. Overall, our findings highlight both the promise and the limitations of LLMs as collaborators in inherently ambiguous qualitative value analysis.

The rapid adoption of vision-language models (VLMs) has heightened the demand for robust intellectual property (IP) protection of these high-value pretrained models. Effective IP protection should proactively confine model deployment within authorized domains and prevent unauthorized transfers. However, existing methods rely on static training-time definitions, limiting flexibility in dynamic environments and often producing opaque responses to unauthorized inputs. To address these limitations, we propose a novel dynamic authorization with legality-aware intellectual property protection (AoD-IP) for VLMs, a framework that supports authorize-on-demand and legality-aware assessment. AoD-IP introduces a lightweight dynamic authorization module that enables flexible, user-controlled authorization, allowing users to actively specify or switch authorized domains on demand at deployment time. This enables the model to adapt seamlessly as application scenarios evolve and provides substantially greater extensibility than existing static-domain approaches. In addition, AoD-IP incorporates a dual-path inference mechanism that jointly predicts input legality-aware and task-specific outputs. Comprehensive experimental results on multiple cross-domain benchmarks demonstrate that AoD-IP maintains strong authorized-domain performance and reliable unauthorized detection, while supporting user-controlled authorization for adaptive deployment in dynamic environments.

Overparameterized ML models, including neural networks, typically induce underdetermined training objectives with multiple global minima. The implicit bias refers to the limiting global minimum that is attained by a common optimization algorithm, such as gradient descent (GD). In this paper, we characterize the implicit bias of GD for training a shallow ReLU model with the squared loss on high-dimensional random features. Prior work showed that the implicit bias does not exist in the worst-case (Vardi and Shamir, 2021), or corresponds exactly to the minimum-l2-norm solution among all global minima under exactly orthogonal data (Boursier et al., 2022). Our work interpolates between these two extremes and shows that, for sufficiently high-dimensional random data, the implicit bias approximates the minimum-l2-norm solution with high probability with a gap on the order $Θ(\sqrt{n/d})$, where n is the number of training examples and d is the feature dimension. Our results are obtained through a novel primal-dual analysis, which carefully tracks the evolution of predictions, data-span coefficients, as well as their interactions, and shows that the ReLU activation pattern quickly stabilizes with high probability over the random data.

Vision-language models are increasingly applied to sensitive domains such as medical imaging and personal photographs, yet existing differentially private methods for in-context learning are limited to few-shot, text-only settings because privacy cost scales with the number of tokens processed. We present Differentially Private Multimodal Task Vectors (DP-MTV), the first framework enabling many-shot multimodal in-context learning with formal $(\varepsilon, δ)$-differential privacy by aggregating hundreds of demonstrations into compact task vectors in activation space. DP-MTV partitions private data into disjoint chunks, applies per-layer clipping to bound sensitivity, and adds calibrated noise to the aggregate, requiring only a single noise addition that enables unlimited inference queries. We evaluate on eight benchmarks across three VLM architectures, supporting deployment with or without auxiliary data. At $\varepsilon=1.0$, DP-MTV achieves 50% on VizWiz compared to 55% non-private and 35% zero-shot, preserving most of the gain from in-context learning under meaningful privacy constraints.

Diverse outputs in text generation are necessary for effective exploration in complex reasoning tasks, such as code generation and mathematical problem solving. Such Pass@$k$ problems benefit from distinct candidates covering the solution space. However, traditional sampling approaches often waste computational resources on repetitive failure modes. While Diffusion Language Models have emerged as a competitive alternative to the prevailing Autoregressive paradigm, they remain susceptible to this redundancy, with independent samples frequently collapsing into similar modes. To address this, we propose a training free, low cost intervention to enhance generative diversity in Diffusion Language Models. Our approach modifies intermediate samples in a batch sequentially, where each sample is repelled from the feature space of previous samples, actively penalising redundancy. Unlike prior methods that require retraining or beam search, our strategy incurs negligible computational overhead, while ensuring that each sample contributes a unique perspective to the batch. We evaluate our method on the HumanEval and GSM8K benchmarks using the LLaDA-8B-Instruct model. Our results demonstrate significantly improved diversity and Pass@$k$ performance across various temperature settings. As a simple modification to the sampling process, our method offers an immediate, low-cost improvement for current and future Diffusion Language Models in tasks that benefit from diverse solution search. We make our code available at https://github.com/sean-lamont/odd.

Context: Open Source Software (OSS) is a crucial component of over 90\% of digital infrastructure underpinning industry and public digital services, facilitating collaborative software development and dissemination. Its significance in the European public sector has been emphasised through various Ministerial Declarations, highlighting its potential to accelerate digitalisation, transform businesses, and foster a digitally skilled population. Research Aim: This study aims to explore how the adoption, development, and collaboration on OSS can be enabled through organisational support functions or centres of competency, also known as Open Source Programme Offices (OSPOs) within Public Sector Organisations (PSOs) in the European Union, Norway, Liechtenstein, and Iceland. Methodology: A qualitative research approach was adopted, involving an interview survey of 18 OSPO representatives across 16 cases of public-sector OSPOs. These cases were cross-analysed and categorised into six OSPO archetypes. The findings were validated and enriched through two follow-up focus groups that included earlier interviewees and additional experts. Results: The study identified six distinct OSPO archetypes, providing insights into their organisational structures, responsibilities, and contributions to OSS adoption. The archetypes, along with policy recommendations, offer guidance on how PSOs can design their own OSPOs, taking into account their specific context, resources, and policy goals. Conclusions: The findings enhance the understanding of OSPOs as strategic endeavours aimed at promoting OSS adoption. The study offers practical guidance for PSOs and policymakers on leveraging OSS to achieve strategic objectives, foster digital sovereignty, drive economic growth, and improve the interoperability and quality of digital services.

Multimodal Federated Learning (MFL) enables clients with heterogeneous data modalities to collaboratively train models without sharing raw data, offering a privacy-preserving framework that leverages complementary cross-modal information. However, existing methods often overlook personalized client performance and struggle with modality/task discrepancies, as well as model heterogeneity. To address these challenges, we propose FedAFD, a unified MFL framework that enhances client and server learning. On the client side, we introduce a bi-level adversarial alignment strategy to align local and global representations within and across modalities, mitigating modality and task gaps. We further design a granularity-aware fusion module to integrate global knowledge into the personalized features adaptively. On the server side, to handle model heterogeneity, we propose a similarity-guided ensemble distillation mechanism that aggregates client representations on shared public data based on feature similarity and distills the fused knowledge into the global model. Extensive experiments conducted under both IID and non-IID settings demonstrate that FedAFD achieves superior performance and efficiency for both the client and the server.

Real-world dialogue usually unfolds as an infinite stream. It thus requires bounded-state memory mechanisms to operate within an infinite horizon. However, existing read-then-think memory is fundamentally misaligned with this setting, as it cannot support ad-hoc memory recall while streams unfold. To explore this challenge, we introduce \textbf{STEM-Bench}, the first benchmark for \textbf{ST}reaming \textbf{E}valuation of \textbf{M}emory. It comprises over 14K QA pairs in dialogue streams that assess perception fidelity, temporal reasoning, and global awareness under infinite-horizon constraints. The preliminary analysis on STEM-Bench indicates a critical \textit{fidelity-efficiency dilemma}: retrieval-based methods use fragment context, while full-context models incur unbounded latency. To resolve this, we propose \textbf{ProStream}, a proactive hierarchical memory framework for streaming dialogues. It enables ad-hoc memory recall on demand by reasoning over continuous streams with multi-granular distillation. Moreover, it employs Adaptive Spatiotemporal Optimization to dynamically optimize retention based on expected utility. It enables a bounded knowledge state for lower inference latency without sacrificing reasoning fidelity. Experiments show that ProStream outperforms baselines in both accuracy and efficiency.

Temporal Forgery Localization (TFL) aims to precisely identify manipulated segments in video and audio, offering strong interpretability for security and forensics. While recent State Space Models (SSMs) show promise in precise temporal reasoning, their use in TFL is hindered by ambiguous boundaries, sparse forgeries, and limited long-range modeling. We propose DeformTrace, which enhances SSMs with deformable dynamics and relay mechanisms to address these challenges. Specifically, Deformable Self-SSM (DS-SSM) introduces dynamic receptive fields into SSMs for precise temporal localization. To further enhance its capacity for temporal reasoning and mitigate long-range decay, a Relay Token Mechanism is integrated into DS-SSM. Besides, Deformable Cross-SSM (DC-SSM) partitions the global state space into query-specific subspaces, reducing non-forgery information accumulation and boosting sensitivity to sparse forgeries. These components are integrated into a hybrid architecture that combines the global modeling of Transformers with the efficiency of SSMs. Extensive experiments show that DeformTrace achieves state-of-the-art performance with fewer parameters, faster inference, and stronger robustness.

Differentially private learning is essential for training models on sensitive data, but empirical studies consistently show that it can degrade performance, introduce fairness issues like disparate impact, and reduce adversarial robustness. The theoretical underpinnings of these phenomena in modern, non-convex neural networks remain largely unexplored. This paper introduces a unified feature-centric framework to analyze the feature learning dynamics of differentially private stochastic gradient descent (DP-SGD) in two-layer ReLU convolutional neural networks. Our analysis establishes test loss bounds governed by a crucial metric: the feature-to-noise ratio (FNR). We demonstrate that the noise required for privacy leads to suboptimal feature learning, and specifically show that: 1) imbalanced FNRs across classes and subpopulations cause disparate impact; 2) even in the same class, noise has a greater negative impact on semantically long-tailed data; and 3) noise injection exacerbates vulnerability to adversarial attacks. Furthermore, our analysis reveals that the popular paradigm of public pre-training and private fine-tuning does not guarantee improvement, particularly under significant feature distribution shifts between datasets. Experiments on synthetic and real-world data corroborate our theoretical findings.

How much can a pitcher's body reveal about the upcoming pitch? We study this question at scale by classifying eight pitch types from monocular 3D pose sequences, without access to ball-flight data. Our pipeline chains a diffusion-based 3D pose backbone with automatic pitching-event detection, groundtruth-validated biomechanical feature extraction, and gradient-boosted classification over 229 kinematic features. Evaluated on 119,561 professional pitches, the largest such benchmark to date, we achieve 80.4\% accuracy using body kinematics alone. A systematic importance analysis reveals that upper-body mechanics contribute 64.9\% of the predictive signal versus 35.1\% for the lower body, with wrist position (14.8\%) and trunk lateral tilt emerging as the most informative joint group and biomechanical feature, respectively. We further show that grip-defined variants (four-seam vs.\ two-seam fastball) are not separable from pose, establishing an empirical ceiling near 80\% and delineating where kinematic information ends and ball-flight information begins.

Accurate time series forecasting underpins decision-making across domains, yet conventional ML development suffers from data scarcity in new deployments, poor adaptability under distribution shift, and diminishing returns from manual iteration. We propose Self-Evolving Agent for Time Series Algorithms (SEA-TS), a framework that autonomously generates, validates, and optimizes forecasting code via an iterative self-evolution loop. Our framework introduces three key innovations: (1) Metric-Advantage Monte Carlo Tree Search (MA-MCTS), which replaces fixed rewards with a normalized advantage score for discriminative search guidance; (2) Code Review with running prompt refinement, where each executed solution undergoes automated review followed by prompt updates that encode corrective patterns, preventing recurrence of similar errors; and (3) Global Steerable Reasoning, which compares each node against global best and worst solutions, enabling cross-trajectory knowledge transfer. We adopt a MAP-Elites archive for architectural diversity. On the public Solar-Energy benchmark, SEA-TS generated code achieves a 40% MAE reduction relative to TimeMixer, surpassing state-of-the-art methods. On proprietary datasets, SEA-TS generated code reduces WAPE by 8.6% on solar PV forecasting and 7.7% on residential load forecasting compared to human-engineered baselines, and achieves 26.17% MAPE on load forecasting versus 29.34% by TimeMixer. Notably, the evolved models discover novel architectural patterns--including physics-informed monotonic decay heads encoding solar irradiance constraints, per-station learned diurnal cycle profiles, and learnable hourly bias correction--demonstrating that autonomous ML engineering can generate genuinely novel algorithmic ideas beyond manual design.

Generating realistic and diverse trajectories is a critical challenge in autonomous driving simulation. While Large Language Models (LLMs) show promise, existing methods often rely on structured data like vectorized maps, which fail to capture the rich, unstructured visual context of a scene. To address this, we propose K-Gen, an interpretable keypoint-guided multimodal framework that leverages Multimodal Large Language Models (MLLMs) to unify rasterized BEV map inputs with textual scene descriptions. Instead of directly predicting full trajectories, K-Gen generates interpretable keypoints along with reasoning that reflects agent intentions, which are subsequently refined into accurate trajectories by a refinement module. To further enhance keypoint generation, we apply T-DAPO, a trajectory-aware reinforcement fine-tuning algorithm. Experiments on WOMD and nuPlan demonstrate that K-Gen outperforms existing baselines, highlighting the effectiveness of combining multimodal reasoning with keypoint-guided trajectory generation.

Preference-based reward learning is widely used for shaping agent behavior to match a user's preference, yet its sparse binary feedback makes it especially vulnerable to causal confusion. The learned reward often latches onto spurious features that merely co-occur with preferred trajectories during training, collapsing when those correlations disappear or reverse at test time. We introduce ReCouPLe, a lightweight framework that uses natural language rationales to provide the missing causal signal. Each rationale is treated as a guiding projection axis in an embedding space, training the model to score trajectories based on features aligned with that axis while de-emphasizing context that is unrelated to the stated reason. Because the same rationales (e.g., "avoids collisions", "completes the task faster") can appear across multiple tasks, ReCouPLe naturally reuses the same causal direction whenever tasks share semantics, and transfers preference knowledge to novel tasks without extra data or language-model fine-tuning. Our learned reward model can ground preferences on the articulated reason, aligning better with user intent and generalizing beyond spurious features. ReCouPLe outperforms baselines by up to 1.5x in reward accuracy under distribution shifts, and 2x in downstream policy performance in novel tasks. We have released our code at https://github.com/mj-hwang/ReCouPLe

Transfer learning is devised to leverage knowledge from pre-trained models to solve new tasks with limited data and computational resources. Meanwhile, dataset distillation has emerged to synthesize a compact dataset that preserves critical information from the original large dataset. Therefore, a combination of transfer learning and dataset distillation offers promising performance in evaluations. However, a non-negligible security threat remains undiscovered in transfer learning using synthetic datasets generated by dataset distillation methods, where an adversary can perform a model hijacking attack with only a few poisoned samples in the synthetic dataset. To reveal this threat, we propose Osmosis Distillation (OD) attack, a novel model hijacking strategy that targets deep learning models using the fewest samples. Comprehensive evaluations on various datasets demonstrate that the OD attack attains high attack success rates in hidden tasks while preserving high model utility in original tasks. Furthermore, the distilled osmosis set enables model hijacking across diverse model architectures, allowing model hijacking in transfer learning with considerable attack performance and model utility. We argue that awareness of using third-party synthetic datasets in transfer learning must be raised.

Instruction following is critical for LLMs deployed in enterprise and API-driven settings, where strict adherence to output formats, content constraints, and procedural requirements is essential for enabling reliable LLM-assisted workflows. However, existing instruction following benchmarks predominantly evaluate natural language generation constraints that reflect the needs of chat assistants rather than enterprise users. To bridge this gap, we introduce FireBench, an LLM instruction following benchmark grounded in real-world enterprise and API usage patterns. FireBench evaluates six core capability dimensions across diverse applications including information extraction, customer support, and coding agents, comprising over 2,400 samples. We evaluate 11 LLMs and present key findings on their instruction following behavior in enterprise scenarios. We open-source FireBench at fire-bench.com to help users assess model suitability, support model developers in diagnosing performance, and invite community contributions.

Student Personas (SPs) are emerging as infrastructure for educational LLMs, yet prior work often relies on ad-hoc prompting or hand-crafted profiles with limited control over educational theory and population distributions. We formalize this as Theory-Aligned and Distribution-Controllable Persona Generation (TAD-PG) and introduce HACHIMI, a multi-agent Propose-Validate-Revise framework that generates theory-aligned, quota-controlled personas. HACHIMI factorizes each persona into a theory-anchored educational schema, enforces developmental and psychological constraints via a neuro-symbolic validator, and combines stratified sampling with semantic deduplication to reduce mode collapse. The resulting HACHIMI-1M corpus comprises 1 million personas for Grades 1-12. Intrinsic evaluation shows near-perfect schema validity, accurate quotas, and substantial diversity, while external evaluation instantiates personas as student agents answering CEPS and PISA 2022 surveys; across 16 cohorts, math and curiosity/growth constructs align strongly between humans and agents, whereas classroom-climate and well-being constructs are only moderately aligned, revealing a fidelity gradient. All personas are generated with Qwen2.5-72B, and HACHIMI provides a standardized synthetic student population for group-level benchmarking and social-science simulations. Resources available at https://github.com/ZeroLoss-Lab/HACHIMI

SinhaLegal introduces a Sinhala legislative text corpus containing approximately 2 million words across 1,206 legal documents. The dataset includes two types of legal documents: 1,065 Acts dated from 1981 to 2014 and 141 Bills from 2010 to 2014, which were systematically collected from official sources. The texts were extracted using OCR with Google Document AI, followed by extensive post-processing and manual cleaning to ensure high-quality, machine-readable content, along with dedicated metadata files for each document. A comprehensive evaluation was conducted, including corpus statistics, lexical diversity, word frequency analysis, named entity recognition, and topic modelling, demonstrating the structured and domain-specific nature of the corpus. Additionally, perplexity analysis using both large and small language models was performed to assess how effectively language models respond to domain-specific texts. The SinhaLegal corpus represents a vital resource designed to support NLP tasks such as summarisation, information extraction, and analysis, thereby bridging a critical gap in Sinhala legal research.

Multi-step theorem prediction is a central challenge in automated reasoning. Existing neural-symbolic approaches rely heavily on supervised parametric models, which exhibit limited generalization to evolving theorem libraries. In this work, we explore training-free theorem prediction through the lens of in-context learning (ICL). We identify a critical scalability bottleneck, termed Structural Drift: as reasoning depth increases, the performance of vanilla ICL degrades sharply, often collapsing to near zero. We attribute this failure to the LLM's inability to recover latent topological dependencies, leading to unstructured exploration. To address this issue, we propose Theorem Precedence Graphs, which encode temporal dependencies from historical solution traces as directed graphs, and impose explicit topological constraints that effectively prune the search space during inference. Coupled with retrieval-augmented graph construction and a stepwise symbolic executor, our approach enables LLMs to act as structured planners without any gradient-based optimization. Experiments on the FormalGeo7k benchmark show that our method achieves 89.29% accuracy, substantially outperforming ICL baselines and matching state-of-the-art supervised models. These results indicate that explicit structural priors offer a promising direction for scaling LLM-based symbolic reasoning.

Why is safety alignment in LLMs shallow? We prove that gradient-based alignment inherently concentrates on positions where harm is decided and vanishes beyond. Using a martingale decomposition of sequence-level harm, we derive an exact characterization of alignment gradients. The gradient at position $t$ equals the covariance between the conditional expected harm and the score function. This implies that positions beyond the harm horizon where the output's harmfulness is already determined receive zero gradient signal during training. This explains empirical observations that KL divergence between aligned and base models concentrates on early tokens. Consequently, standard alignment objectives cannot produce deep alignment, regardless of optimization quality. We introduce the concept of harm information $I_t$, which quantifies each position's influence on harm, and prove that equilibrium KL divergence tracks this quantity. Finally, we derive an objective based on recovery penalties that creates gradient signal at all positions, providing theoretical grounding for empirically successful data augmentation techniques.

Speech production is a complex process spanning neural planning, motor control, muscle activation, and articulatory kinematics. While the acoustic speech signal is the most accessible product of the speech production act, it does not directly reveal its causal neurophysiological substrates. We present the first simultaneous acquisition of real-time (dynamic) MRI, EEG, and surface EMG, capturing several key aspects of the speech production chain: brain signals, muscle activations, and articulatory movements. This multimodal acquisition paradigm presents substantial technical challenges, including MRI-induced electromagnetic interference and myogenic artifacts. To mitigate these, we introduce an artifact suppression pipeline tailored to this tri-modal setting. Once fully developed, this framework is poised to offer an unprecedented window into speech neuroscience and insights leading to brain-computer interface advances.

We introduce the Dynamic Behavioral Constraint (DBC) benchmark, the first empirical framework for evaluating the efficacy of a structured, 150-control behavioral governance layer, the MDBC (Madan DBC) system, applied at inference time to large language models (LLMs). Unlike training time alignment methods (RLHF, DPO) or post-hoc content moderation APIs, DBCs constitute a system prompt level governance layer that is model-agnostic, jurisdiction-mappable, and auditable. We evaluate the DBC Framework across a 30 domain risk taxonomy organized into six clusters (Hallucination and Calibration, Bias and Fairness, Malicious Use, Privacy and Data Protection, Robustness and Reliability, and Misalignment Agency) using an agentic red-team protocol with five adversarial attack strategies (Direct, Roleplay, Few-Shot, Hypothetical, Authority Spoof) across 3 model families. Our three-arm controlled design (Base, Base plus Moderation, Base plus DBC) enables causal attribution of risk reduction. Key findings: the DBC layer reduces the aggregate Risk Exposure Rate (RER) from 7.19 percent (Base) to 4.55 percent (Base plus DBC), representing a 36.8 percent relative risk reduction, compared with 0.6 percent for a standard safety moderation prompt. MDBC Adherence Scores improve from 8.6 by 10 (Base) to 8.7 by 10 (Base plus DBC). EU AI Act compliance (automated scoring) reaches 8.5by 10 under the DBC layer. A three judge evaluation ensemble yields Fleiss kappa greater than 0.70 (substantial agreement), validating our automated pipeline. Cluster ablation identifies the Integrity Protection cluster (MDBC 081 099) as delivering the highest per domain risk reduction, while graybox adversarial attacks achieve a DBC Bypass Rate of 4.83 percent . We release the benchmark code, prompt database, and all evaluation artefacts to enable reproducibility and longitudinal tracking as models evolve.

Communication can improve coordination in partially observed multi-agent reinforcement learning (MARL), but learning \emph{when} and \emph{who} to communicate with requires choosing among many possible sender-recipient pairs, and the effect of any single message on future reward is hard to isolate. We introduce \textbf{SCoUT} (\textbf{S}calable \textbf{Co}mmunication via \textbf{U}tility-guided \textbf{T}emporal grouping), which addresses both these challenges via temporal and agent abstraction within traditional MARL. During training, SCoUT resamples \textit{soft} agent groups every \(K\) environment steps (macro-steps) via Gumbel-Softmax; these groups are latent clusters that induce an affinity used as a differentiable prior over recipients. Using the same assignments, a group-aware critic predicts values for each agent group and maps them to per-agent baselines through the same soft assignments, reducing critic complexity and variance. Each agent is trained with a three-headed policy: environment action, send decision, and recipient selection. To obtain precise communication learning signals, we derive counterfactual communication advantages by analytically removing each sender's contribution from the recipient's aggregated messages. This counterfactual computation enables precise credit assignment for both send and recipient-selection decisions. At execution time, all centralized training components are discarded and only the per-agent policy is run, preserving decentralized execution. Project website, videos and code: \hyperlink{https://scout-comm.github.io/}{https://scout-comm.github.io/}

Popular explanation methods often produce unreliable feature importance scores due to missingness bias, a systematic distortion that arises when models are probed with ablated, out-of-distribution inputs. Existing solutions treat this as a deep representational flaw that requires expensive retraining or architectural modifications. In this work, we challenge this assumption and show that missingness bias can be effectively treated as a superficial artifact of the model's output space. We introduce MCal, a lightweight post-hoc method that corrects this bias by fine-tuning a simple linear head on the outputs of a frozen base model. Surprisingly, we find this simple correction consistently reduces missingness bias and is competitive with, or even outperforms, prior heavyweight approaches across diverse medical benchmarks spanning vision, language, and tabular domains.

Pre-training data detection for LLMs is essential for addressing copyright concerns and mitigating benchmark contamination. Existing methods mainly focus on the likelihood-based statistical features or heuristic signals before and after fine-tuning, but the former are susceptible to word frequency bias in corpora, and the latter strongly depend on the similarity of fine-tuning data. From an optimization perspective, we observe that during training, samples transition from unfamiliar to familiar in a manner reflected by systematic differences in gradient behavior. Familiar samples exhibit smaller update magnitudes, distinct update locations in model components, and more sharply activated neurons. Based on this insight, we propose GDS, a method that identifies pre-training data by probing Gradient Deviation Scores of target samples. Specifically, we first represent each sample using gradient profiles that capture the magnitude, location, and concentration of parameter updates across FFN and Attention modules, revealing consistent distinctions between member and non-member data. These features are then fed into a lightweight classifier to perform binary membership inference. Experiments on five public datasets show that GDS achieves state-of-the-art performance with significantly improved cross-dataset transferability over strong baselines. Further interpretability analyse show gradient feature distribution differences, enabling practical and scalable pre-training data detection.

Algorithmic speedup of training common neural architectures is made difficult by the lack of structure guaranteed by the function compositions inherent to such networks. In contrast to multilayer perceptrons (MLPs), Kolmogorov-Arnold networks (KANs) provide more structure by expanding learned activations in a specified basis. This paper exploits this structure to develop practical algorithms and theoretical insights, yielding training speedup via multilevel training for KANs. To do so, we first establish an equivalence between KANs with spline basis functions and multichannel MLPs with power ReLU activations through a linear change of basis. We then analyze how this change of basis affects the geometry of gradient-based optimization with respect to spline knots. The KANs change-of-basis motivates a multilevel training approach, where we train a sequence of KANs naturally defined through a uniform refinement of spline knots with analytic geometric interpolation operators between models. The interpolation scheme enables a ``properly nested hierarchy'' of architectures, ensuring that interpolation to a fine model preserves the progress made on coarse models, while the compact support of spline basis functions ensures complementary optimization on subsequent levels. Numerical experiments demonstrate that our multilevel training approach can achieve orders of magnitude improvement in accuracy over conventional methods to train comparable KANs or MLPs, particularly for physics informed neural networks. Finally, this work demonstrates how principled design of neural networks can lead to exploitable structure, and in this case, multilevel algorithms that can dramatically improve training performance.

This is the final paper in the five-part series The Semantic Arrow of Time. Part I identified the FITO category mistake -- treating forward temporal flow as sufficient for establishing meaning. Part II presented the constructive alternative: the OAE link state machine with its mandatory reflecting phase. Part III showed the FITO fallacy operating at industrial scale in RDMA completion semantics. Part IV traced the same pattern through file synchronization, email, human memory, and language model hallucination. This paper closes the series by constructing the Leibniz Bridge: a unified framework that connects the philosophical foundations (Leibniz's Identity of Indiscernibles, as formalized by Spekkens), the protocol engineering (OAE's bilateral transaction structure), and the physical substrate (indefinite causal order in quantum mechanics). The bridge rests on a single principle: mutual information conservation -- the requirement that every causal exchange preserve the total information accessible to both endpoints, with the direction of time emerging not from axiom but from entropy production when a reversible exchange commits. We show that this principle dissolves the apparent impossibility of the FLP, Two Generals, and CAP theorems by revealing them as theorems about FITO systems, not about physics. We present the triangle network as the minimal topology for semantic consistency without centralized coordination. We conclude with open questions and a reflection on what distributed computing looks like when the FITO assumption is dropped.

Partial label learning is a prominent weakly supervised classification task, where each training instance is ambiguously labeled with a set of candidate labels. In real-world scenarios, candidate labels are often influenced by instance features, leading to the emergence of instance-dependent PLL (ID-PLL), a setting that more accurately reflects this relationship. A significant challenge in ID-PLL is instance entanglement, where instances from similar classes share overlapping features and candidate labels, resulting in increased class confusion. To address this issue, we propose a novel Class-specific Augmentation based Disentanglement (CAD) framework, which tackles instance entanglement by both intra- and inter-class regulations. For intra-class regulation, CAD amplifies class-specific features to generate class-wise augmentations and aligns same-class augmentations across instances. For inter-class regulation, CAD introduces a weighted penalty loss function that applies stronger penalties to more ambiguous labels, encouraging larger inter-class distances. By jointly applying intra- and inter-class regulations, CAD improves the clarity of class boundaries and reduces class confusion caused by entanglement. Extensive experimental results demonstrate the effectiveness of CAD in mitigating the entanglement problem and enhancing ID-PLL performance. The code is available at https://github.com/RyanZhaoIc/CAD.git.

Aligning Large Language Models (LLMs) with nuanced human values remains a critical challenge, as existing methods like Reinforcement Learning from Human Feedback (RLHF) often handle only coarse-grained attributes. In practice, fine-tuning LLMs on task-specific datasets to optimize value alignment inevitably incurs an alignment tax: the model's pre-calibrated value system drifts significantly due to latent bias absorption from training data, while the fine-tuning process also causes severe hallucinations and semantic information loss in generated responses. To address this, we propose VISA (Value Injection via Shielded Adaptation), a closed-loop framework designed to navigate this trade-off. VISA's architecture features a high-precision value detector, a semantic-to-value translator, and a core value-rewriter. The value-rewriter is trained via Group Relative Policy Optimization (GRPO) with a composite reward function that simultaneously optimizes for fine-grained value precision, and the preservation of semantic integrity. By learning an optimal policy to balance these competing objectives, VISA effectively mitigates the alignment tax while staying loyal to the original knowledge. Our experiments demonstrate that this approach enables precise control over a model's value expression while maintaining its factual consistency and general capabilities, significantly outperforming both standard fine-tuning methods and prompting-based baselines, including GPT-4o.

Automated short-answer scoring lags other LLM applications. We meta-analyze 890 culminating results across a systematic review of LLM short-answer scoring studies, modeling the traditional effect size of Quadratic Weighted Kappa (QWK) with mixed effects metaregression. We quantitatively illustrate that that the level of difficulty for human experts to perform the task of scoring written work of children has no observed statistical effect on LLM performance. Particularly, we show that some scoring tasks measured as the easiest by human scorers were the hardest for LLMs. Whether by poor implementation by thoughtful researchers or patterns traceable to autoregressive training, on average decoder-only architectures underperform encoders by 0.37--a substantial difference in agreement with humans. Additionally, we measure the contributions of various aspects of LLM technology on successful scoring such as tokenizer vocabulary size, which exhibits diminishing returns--potentially due to undertrained tokens. Findings argue for systems design which better anticipates known statistical shortcomings of autoregressive models. Finally, we provide additional experiments to illustrate wording and tokenization sensitivity and bias elicitation in high-stakes education contexts, where LLMs demonstrate racial discrimination. Code and data for this study are available.

Embodied foundation models are increasingly performant in real-world domains such as robotics or autonomous driving. These models are often deployed in interactive or assistive settings, where it is important that these assistive models generalize to new users and new tasks. Diverse interactive data generation offers a promising avenue for providing data-efficient generalization capabilities for interactive embodied foundation models. In this paper, we investigate the generalization capabilities of a multimodal foundation model fine-tuned on diverse interactive assistance data in a synthetic domain. We explore generalization along two axes: a) assistance with unseen categories of user behavior and b) providing guidance in new configurations not encountered during training. We study a broad capability called \textbf{Open-Set Corrective Assistance}, in which the model needs to inspect lengthy user behavior and provide assistance through either corrective actions or language-based feedback. This task remains unsolved in prior work, which typically assumes closed corrective categories or relies on external planners, making it a challenging testbed for evaluating the limits of assistive data. To support this task, we generate synthetic assistive datasets in Overcooked and fine-tune a LLaMA-based model to evaluate generalization to novel tasks and user behaviors. Our approach provides key insights into the nature of assistive datasets required to enable open-set assistive intelligence. In particular, we show that performant models benefit from datasets that cover different aspects of assistance, including multimodal grounding, defect inference, and exposure to diverse scenarios.

Port congestion at major maritime hubs disrupts global supply chains, yet existing prediction systems typically prioritize forecasting accuracy without providing operationally interpretable explanations. This paper proposes AIS-TGNN, an evidence-grounded framework that jointly performs congestion-escalation prediction and faithful natural-language explanation by coupling a Temporal Graph Attention Network (TGAT) with a structured large language model (LLM) reasoning module. Daily spatial graphs are constructed from Automatic Identification System (AIS) broadcasts, where each grid cell represents localized vessel activity and inter-cell interactions are modeled through attention-based message passing. The TGAT predictor captures spatiotemporal congestion dynamics, while model-internal evidence, including feature z-scores and attention-derived neighbor influence, is transformed into structured prompts that constrain LLM reasoning to verifiable model outputs. To evaluate explanatory reliability, we introduce a directional-consistency validation protocol that quantitatively measures agreement between generated narratives and underlying statistical evidence. Experiments on six months of AIS data from the Port of Los Angeles and Long Beach demonstrate that the proposed framework outperforms both LR and GCN baselines, achieving a test AUC of 0.761, AP of 0.344, and recall of 0.504 under a strict chronological split while producing explanations with 99.6% directional consistency. Results show that grounding LLM generation in graph-model evidence enables interpretable and auditable risk reporting without sacrificing predictive performance. The framework provides a practical pathway toward operationally deployable explainable AI for maritime congestion monitoring and supply-chain risk management.

Manipulative communication, such as gaslighting, guilt-tripping, and emotional coercion, is often difficult for individuals to recognize. Existing agentic AI systems lack the structured, longitudinal memory to track these subtle, context-dependent tactics, often failing due to limited context windows and catastrophic forgetting. We introduce EchoGuard, an agentic AI framework that addresses this gap by using a Knowledge Graph (KG) as the agent's core episodic and semantic memory. EchoGuard employs a structured Log-Analyze-Reflect loop: (1) users log interactions, which the agent structures as nodes and edges in a personal, episodic KG (capturing events, emotions, and speakers); (2) the system executes complex graph queries to detect six psychologically-grounded manipulation patterns (stored as a semantic KG); and (3) an LLM generates targeted Socratic prompts grounded by the subgraph of detected patterns, guiding users toward self-discovery. This framework demonstrates how the interplay between agentic architectures and Knowledge Graphs can empower individuals in recognizing manipulative communication while maintaining personal autonomy and safety. We present the theoretical foundation, framework design, a comprehensive evaluation strategy, and a vision to validate this approach.

Persistent conversational AI systems face a choice between passing full conversation histories to a long-context large language model (LLM) and maintaining a dedicated memory system that extracts and retrieves structured facts. We compare a fact-based memory system built on the Mem0 framework against long-context LLM inference on three memory-centric benchmarks - LongMemEval, LoCoMo, and PersonaMemv2 - and evaluate both architectures on accuracy and cumulative API cost. Long-context GPT-5-mini achieves higher factual recall on LongMemEval and LoCoMo, while the memory system is competitive on PersonaMemv2, where persona consistency depends on stable, factual attributes suited to flat-typed extraction. We construct a cost model that incorporates prompt caching and show that the two architectures have structurally different cost profiles: long-context inference incurs a per-turn charge that grows with context length even under caching, while the memory system's per-turn read cost remains roughly fixed after a one-time write phase. At a context length of 100k tokens, the memory system becomes cheaper after approximately ten interaction turns, with the break-even point decreasing as context length grows. These results characterize the accuracy-cost trade-off between the two approaches and provide a concrete criterion for selecting between them in production deployments.

Polarity is a fundamental reciprocal duality of $n$-dimensional projective geometry which associates to points polar hyperplanes, and more generally $k$-dimensional convex bodies to polar $(n-1-k)$-dimensional convex bodies. It is well-known that the Legendre-Fenchel transformation of functions can be interpreted from the polarity viewpoint of their graphs using an extra dimension. In this paper, we first show that generic polarities induced by quadratic polarity functionals can be expressed either as deformed Legendre polarity or as the Legendre polarity of deformed convex bodies, and be efficiently manipulated using linear algebra on $(n+2)\times (n+2)$ matrices operating on homogeneous coordinates. Second, we define polar divergences using the Legendre polarity and show that they generalize the Fenchel-Young divergence or equivalent Bregman divergence. This polarity study brings new understanding of the core reference duality in information geometry. Last, we show that the total Bregman divergences can be considered as a total polar Fenchel-Young divergence from which we newly exhibit the reference duality using dual polar conformal factors.

We present Meta-D, an architecture that explicitly leverages categorical scanner metadata such as MRI sequence and plane orientation to guide feature extraction for brain tumor analysis. We aim to improve the performance of medical image deep learning pipelines by integrating explicit metadata to stabilize feature representations. We first evaluate this in 2D tumor detection, where injecting sequence (e.g., T1, T2) and plane (e.g., axial) metadata dynamically modulates convolutional features, yielding an absolute increase of up to 2.62% in F1-score over image-only baselines. Because metadata grounds feature extraction when data are available, we hypothesize it can serve as a robust anchor when data are missing. We apply this to 3D missing-modality tumor segmentation. Our Transformer Maximizer utilizes metadata-based cross-attention to isolate and route available modalities, ensuring the network focuses on valid slices. This targeted attention improves brain tumor segmentation Dice scores by up to 5.12% under extreme modality scarcity while reducing model parameters by 24.1%.

This is the fourth of five papers comprising The Semantic Arrow of Time. Parts I-III established that computing's hidden arrow of time is semantic rather than thermodynamic, that bilateral transaction protocols create causal order through a mandatory reflecting phase, and that RDMA's completion semantics implement the FITO category mistake at industrial scale. This paper traces the consequences of the FITO category mistake beyond the data center, into systems people use every day. We examine three domains where forward-only temporal assumptions destroy meaning: file synchronization, where cloud platforms silently delete user content because last-writer-wins cannot represent distributed causality; email, where timestamp-based ordering produces phantom messages, causality violations, and stuck synchronization; and memory--both human and artificial--where reconstructive processes that operate without transactional guarantees produce systematic semantic corruption. In each domain, we identify the same structural pattern: a system that commits state changes forward in time without a reflecting phase, and that therefore cannot distinguish between successful semantic integration and mere temporal succession. The pattern is not coincidental. It is the FITO category mistake operating at different scales: bytes in a NIC buffer, files in a cloud, messages in an inbox, engrams in a hippocampus, tokens in a transformer. We conclude that the semantic arrow of time is violated whenever a system treats the forward flow of information as sufficient evidence of meaning. Part V will show how the Leibniz Bridge provides a unified framework for closing this gap across all five domains.

This paper presents our solution for the DL Sprint 4.0, addressing the dual challenges of Bengali Long-Form Speech Recognition (Task 1) and Speaker Diarization (Task 2). Processing long-form, multi-speaker Bengali audio introduces significant hurdles in voice activity detection, overlapping speech, and context preservation. To solve the long-form transcription challenge, we implemented a robust audio chunking strategy utilizing whisper-timestamped, allowing us to feed precise, context-aware segments into our fine-tuned acoustic model for high-accuracy transcription. For the diarization task, we developed an integrated pipeline leveraging pyannote.audio and WhisperX. A key contribution of our approach is the domain-specific fine-tuning of the Pyannote segmentation model on the competition dataset. This adaptation allowed the model to better capture the nuances of Bengali conversational dynamics and accurately resolve complex, overlapping speaker boundaries. Our methodology demonstrates that applying intelligent timestamped chunking to ASR and targeted segmentation fine-tuning to diarization significantly drives down Word Error Rate (WER) and Diarization Error Rate (DER), in low-resource settings.

We study how architectural inductive bias reshapes the implicit regularization induced by the edge-of-stability phenomenon in gradient descent. Prior work has established that for fully connected networks, the strength of this regularization is governed solely by the global input geometry; consequently, it is insufficient to prevent overfitting on difficult distributions such as the high-dimensional sphere. In this paper, we show that locality and weight sharing fundamentally change this picture. Specifically, we prove that provided the receptive field size $m$ remains small relative to the ambient dimension $d$, these networks generalize on spherical data with a rate of $n^{-\frac{1}{6} +O(m/d)}$, a regime where fully connected networks provably fail. This theoretical result confirms that weight sharing couples the learned filters to the low-dimensional patch manifold, thereby bypassing the high dimensionality of the ambient space. We further corroborate our theory by analyzing the patch geometry of natural images, showing that standard convolutional designs induce patch distributions that are highly amenable to this stability mechanism, thus providing a systematic explanation for the superior generalization of convolutional networks over fully connected baselines.

The limited understanding capacity of the visual encoder in Contrastive Language-Image Pre-training (CLIP) has become a key bottleneck for downstream performance. This capacity includes both Discriminative Ability (D-Ability), which reflects class separability, and Detail Perceptual Ability (P-Ability), which focuses on fine-grained visual cues. Recent solutions use diffusion models to enhance representations by conditioning image reconstruction on CLIP visual tokens. We argue that such paradigms may compromise D-Ability and therefore fail to effectively address CLIP's representation limitations. To address this, we integrate contrastive signals into diffusion-based reconstruction to pursue more comprehensive visual representations. We begin with a straightforward design that augments the diffusion process with contrastive learning on input images. However, empirical results show that the naive combination suffers from gradient conflict and yields suboptimal performance. To balance the optimization, we introduce the Diffusion Contrastive Reconstruction (DCR), which unifies the learning objective. The key idea is to inject contrastive signals derived from each reconstructed image, rather than from the original input, into the diffusion process. Our theoretical analysis shows that the DCR loss can jointly optimize D-Ability and P-Ability. Extensive experiments across various benchmarks and multi-modal large language models validate the effectiveness of our method. The code is available at https://github.com/boyuh/DCR.

Electromagnetic (EM) shielding is widely used to suppress radiated emissions and limit passive EM side-channel leakage. However, shielding does not address active probing, where an adversary injects external radio-frequency (RF) signals and observes the device's reflective response. This work studies whether such impedance-modulated backscattering persists when radiated emissions are suppressed by shielding. By injecting controlled RF signals and analyzing the reflections, we demonstrate that state-dependent impedance variations remain observable at frequencies outside the shields' primary attenuation band. Using processors implemented on FPGA and microcontroller prototypes, and evaluating workload profiles under three industry-standard shields, we find that passive EM measurements lose discriminative power under shielding, while backscattering responses remain separable. These results indicate that active RF probing can expose execution-dependent behavior even in shielded systems, motivating the need to consider active impedance-based probing within hardware security evaluation flows.

Large language models (LLMs) are increasingly used for semantic query processing over large corpora. A set of semantic operators derived from relational algebra has been proposed to provide a unified interface for expressing such queries, among which the semantic filter operator serves as a cornerstone. Given a table T with a natural language predicate e, for each tuple in the relation, the execution of a semantic filter proceeds by constructing an input prompt that combines the predicate e with its content, querying the LLM, and obtaining the binary decision. However, this tuple-by-tuple evaluation necessitates a complete linear scan of the table, incurring prohibitive latency and token costs. Although recent work has attempted to optimize semantic filtering, it still does not break the linear LLM invocation barriers. To address this, we propose Clustering-Sampling-Voting (CSV), a new framework that reduces LLM invocations to sublinear complexity while providing error guarantees. CSV embeds tuples into semantic clusters, samples a small subset for LLM evaluation, and infers cluster-level labels via two proposed voting strategies: UniVote, which aggregates labels uniformly, and SimVote, which weights votes by semantic similarity. Moreover, CSV triggers re-clustering on ambiguous clusters to ensure robustness across diverse datasets. The results conducted on real-world datasets demonstrate that CSV reduces the number of LLM calls by 1.28-355x compared to the state-of-the-art approaches, while maintaining comparable effectiveness in terms of Accuracy and F1 score.

Large language models (LLMs) have been widely deployed for online generative services, where numerous LLM instances jointly handle workloads with fluctuating request arrival rates and variable request lengths. To efficiently execute coexisting compute-intensive and memory-intensive operators, near-memory processing (NMP) based computing paradigm has been extensively proposed. However, existing NMP designs adopt coarse-grained KV cache management and inflexible attention execution flow. Such limitations hinder these proposals from efficiently handling \textit{highly dynamic} LLM serving workloads, limiting their ability to accelerate LLM serving. To tackle these problems, we propose Helios, a Hybrid-bonding-based \uline{L}LM \uline{S}erving accelerator. Helios aims to bridge the fundamental gap between the dynamic nature of KV cache management in LLM serving and the distributed, non-uniform memory abstraction among NMP processing engines (PEs). To this end, we design both the intra-PE execution flow and the inter-PE communication primitives for distributed tiled attention execution. We further propose \textit{spatially-aware} KV cache allocation mechanism to balance the attention workload distribution while minimizing the inter-PE data transfer overhead. Compared with existing GPU/NMP designs, Helios achieves 3.25 times (geomean) speedup and 3.36 times (geomean) better energy efficiency, along with up to 72%/76% P50/P99 time-between-tokens degradation.

Segmentation is crucial for brain gliomas as it delineates the glioma s extent and location, aiding in precise treatment planning and monitoring, thus improving patient outcomes. Accurate segmentation ensures proper identification of the glioma s size and position, transforming images into applicable data for analysis. Classification of brain gliomas is also essential because different types require different treatment approaches. Accurately classifying brain gliomas by size, location, and aggressiveness is essential for personalized prognosis prediction, follow-up care, and monitoring disease progression, ensuring effective diagnosis, treatment, and management. In glioma research, irregular tissues are often observable, but error free and reproducible segmentation is challenging. Many researchers have surveyed brain glioma segmentation, proposing both fully automatic and semi-automatic techniques. The adoption of these methods by radiologists depends on ease of use and supervision, with semi-automatic techniques preferred due to the need for accurate evaluations. This review evaluates effective segmentation and classification techniques post magnetic resonance imaging acquisition, highlighting that convolutional neural network architectures outperform traditional techniques in these tasks.

Medical image analysis relies on accurate segmentation, and benefits from controllable synthesis (of new training images). Yet both tasks of the cyclical pipeline face spatial imbalance: lesions occupy small regions against vast backgrounds. In particular, diffusion models have been shown to drift from prescribed lesion layouts, while efficient segmenters struggle on spatially uncertain regions. Adaptive spatial weighting addresses this by learning where to allocate computational resources. This paper introduces a pair of network adapters: 1) Learnable Adaptive Weighter (LAW) which predicts per-pixel loss modulation from features and masks for diffusion training, stabilized via a mix of normalization, clamping, and regularization to prevent degenerate solutions; and 2) Optimal Region Detection with Efficient Resolution (ORDER) which applies selective bidirectional skip attention at late decoder stages for efficient segmentation. Experiments on polyp and kidney tumor datasets demonstrate that LAW achieves 20% FID generative improvement over a uniform baseline (52.28 vs. 65.60), with synthetic data then improving downstream segmentation by 4.9% Dice coefficient (83.2% vs. 78.3%). ORDER reaches 6.0% Dice improvement on MK-UNet (81.3% vs. 75.3%) with 0.56 GFLOPs and just 42K parameters, remaining 730x smaller than the standard nnUNet.

We introduce Timer-S1, a strong Mixture-of-Experts (MoE) time series foundation model with 8.3B total parameters, 0.75B activated parameters for each token, and a context length of 11.5K. To overcome the scalability bottleneck in existing pre-trained time series foundation models, we perform Serial Scaling in three dimensions: model architecture, dataset, and training pipeline. Timer-S1 integrates sparse TimeMoE blocks and generic TimeSTP blocks for Serial-Token Prediction (STP), a generic training objective that adheres to the serial nature of forecasting. The proposed paradigm introduces serial computations to improve long-term predictions while avoiding costly rolling-style inference and pronounced error accumulation in the standard next-token prediction. Pursuing a high-quality and unbiased training dataset, we curate TimeBench, a corpus with one trillion time points, and apply meticulous data augmentation to mitigate predictive bias. We further pioneer a post-training stage, including continued pre-training and long-context extension, to enhance short-term and long-context performance. Evaluated on the large-scale GIFT-Eval leaderboard, Timer-S1 achieves state-of-the-art forecasting performance, attaining the best MASE and CRPS scores as a pre-trained model. Timer-S1 will be released to facilitate further research.

Reinforcement learning (RL) has been extensively employed in a wide range of decision-making problems, such as games and robotics. Recently, diffusion policies have shown strong potential in modeling multi-modal behaviors, enabling more diverse and flexible action generation compared to the conventional Gaussian policy. Despite various attempts to combine RL with diffusion, a key challenge is the difficulty of computing action log-likelihood under the diffusion model. This greatly hinders the direct application of diffusion policies in on-policy reinforcement learning. Most existing methods calculate or approximate the log-likelihood through the entire denoising process in the diffusion model, which can be memory- and computationally inefficient. To overcome this challenge, we propose a novel and efficient method to train a diffusion policy in an on-policy setting that requires only evaluating a simple Gaussian probability. This is achieved by aligning the policy iteration with the diffusion process, which is a distinct paradigm compared to previous work. Moreover, our formulation can naturally handle entropy regularization, which is often difficult to incorporate into diffusion policies. Experiments demonstrate that the proposed method produces multimodal policy behaviors and achieves superior performance on a variety of benchmark tasks in both IsaacLab and MuJoCo Playground.

While LLMs demonstrate strong reasoning capabilities when provided with full information in a single turn, they exhibit substantial vulnerability in multi-turn interactions. Specifically, when information is revealed incrementally or requires updates, models frequently fail to integrate new constraints, leading to a collapse in performance compared to their single-turn baselines. We term the root cause as \emph{Contextual Inertia}: a phenomenon where models rigidly adhere to previous reasoning traces. Even when users explicitly provide corrections or new data in later turns, the model ignores them, preferring to maintain consistency with its previous (incorrect) reasoning path. To address this, we introduce \textbf{R}einforcement \textbf{L}earning with \textbf{S}ingle-\textbf{T}urn \textbf{A}nchors (\textbf{RLSTA}), a generalizable training approach designed to stabilize multi-turn interaction across diverse scenarios and domains. RLSTA leverages the model's superior single-turn capabilities as stable internal anchors to provide reward signals. By aligning multi-turn responses with these anchors, RLSTA empowers models to break contextual inertia and self-calibrate their reasoning based on the latest information. Experiments show that RLSTA significantly outperforms standard fine-tuning and abstention-based methods. Notably, our method exhibits strong cross-domain generalization (e.g., math to code) and proves effective even without external verifiers, highlighting its potential for general-domain applications.

Python's Global Interpreter Lock prevents execution on more than one CPU core at the same time, even when multiple threads are used. However, starting with Python 3.13 an experimental build allows disabling the GIL. While prior work has examined speedup implications of this disabling, the effects on energy consumption and hardware utilization have received less attention. This study measures execution time, CPU utilization, memory usage, and energy consumption using four workload categories: NumPy-based, sequential kernels, threaded numerical workloads, and threaded object workloads, comparing GIL and free-threaded builds of Python 3.14.2. The results highlight a trade-off. For parallelizable workloads operating on independent data, the free-threaded build reduces execution time by up to 4 times, with a proportional reduction in energy consumption, and effective multi-core utilization, at the cost of an increase in memory usage. In contrast, sequential workloads do not benefit from removing the GIL and instead show a 13-43% increase in energy consumption. Similarly, workloads where threads frequently access and modify the same objects show reduced improvements or even degradation due to lock contention. Across all workloads, energy consumption is proportional to execution time, indicating that disabling the GIL does not significantly affect power consumption, even when CPU utilization increases. When it comes to memory, the no-GIL build shows a general increase, more visible in virtual memory than in physical memory. This increase is primarily attributed to per-object locking, additional thread-safety mechanisms in the runtime, and the adoption of a new memory allocator. These findings suggest that Python's no-GIL build is not a universal improvement. Developers should evaluate whether their workload can effectively benefit from parallel execution before adoption.

Causal discovery with latent variables is a fundamental task. Yet most existing methods rely on strong structural assumptions, such as enforcing specific indicator patterns for latents or restricting how they can interact with others. We argue that a core obstacle to a general, structural-assumption-free approach is the lack of an equivalence characterization: without knowing what can be identified, one generally cannot design methods for how to identify it. In this work, we aim to close this gap for linear non-Gaussian models. We establish the graphical criterion for when two graphs with arbitrary latent structure and cycles are distributionally equivalent, that is, they induce the same observed distribution set. Key to our approach is a new tool, edge rank constraints, which fills a missing piece in the toolbox for latent-variable causal discovery in even broader settings. We further provide a procedure to traverse the whole equivalence class and develop an algorithm to recover models from data up to such equivalence. To our knowledge, this is the first equivalence characterization with latent variables in any parametric setting without structural assumptions, and hence the first structural-assumption-free discovery method. Code and an interactive demo are available at https://equiv.cc.

With the increasing deployment of intelligent sensing technologies in highly sensitive environments such as restrooms and locker rooms, visual surveillance systems face a profound privacy-security paradox. Existing privacy-preserving approaches, including physical desensitization, encryption, and obfuscation, often compromise semantic understanding or fail to ensure mathematically provable irreversibility. Although Privacy Camera 1.0 eliminated visual data at the source to prevent leakage, it provided only textual judgments, leading to evidentiary blind spots in disputes. To address these limitations, this paper proposes a novel privacy-preserving perception framework based on the AI Flow paradigm and a collaborative edge-cloud architecture. By deploying a visual desensitizer at the edge, raw images are transformed in real time into abstract feature vectors through nonlinear mapping and stochastic noise injection under the Information Bottleneck principle, ensuring identity-sensitive information is stripped and original images are mathematically unreconstructable. The abstract representations are transmitted to the cloud for behavior recognition and semantic reconstruction via a "dynamic contour" visual language, achieving a critical balance between perception and privacy while enabling illustrative visual reference without exposing raw images.

This is the third of five papers comprising The Semantic Arrow of Time. Parts I and II identified computing's hidden semantic arrow of time, the FITO category mistake, and presented the constructive alternative: the OAE link state machine with its mandatory reflecting phase. This paper examines what happens when those principles are violated at industrial scale. Remote Direct Memory Access (RDMA) is the highest-performance data movement technology in production, deployed across Meta's 24,000-GPU clusters, Google's data centers, and Microsoft's Azure infrastructure. We argue that RDMA's completion semantics contain a category mistake: they guarantee placement (data written to a remote NIC buffer) but not commitment (data semantically integrated by the receiving application). We call this the completion fallacy. We document the fallacy through seven temporal stages of an RDMA Write operation, showing that the gap between completion signal and application semantic satisfaction can be arbitrarily large. We trace consequences through four case studies: Meta's RoCE fabric, Google's 1RMA redesign, Microsoft's DCQCN failures, and SDR-RDMA partial completions. A comparative analysis shows CXL 3.0, NVLink, and UALink each address parts of the completion fallacy but none eliminates it entirely. Only a protocol architecture with a mandatory reflecting phase can close the gap between delivery and commitment.

Despite the exceptional reasoning capabilities of Multimodal Large Language Models (MLLMs), their adaptation into universal embedding models is significantly impeded by task conflict. To address this, we propose TSEmbed, a universal multimodal embedding framework that synergizes Mixture-of-Experts (MoE) with Low-Rank Adaptation (LoRA) to explicitly disentangle conflicting task objectives. Moreover, we introduce Expert-Aware Negative Sampling (EANS), a novel strategy that leverages expert routing distributions as an intrinsic proxy for semantic similarity. By dynamically prioritizing informative hard negatives that share expert activation patterns with the query, EANS effectively sharpens the model's discriminative power and refines embedding boundaries. To ensure training stability, we further devise a two-stage learning paradigm that solidifies expert specialization before optimizing representations via EANS. TSEmbed achieves state-of-the-art performance on both the Massive Multimodal Embedding Benchmark (MMEB) and real-world industrial production datasets, laying a foundation for task-level scaling in universal multimodal embeddings.

Dental crown restoration is one of the most common treatment modalities for tooth defect, where personalized dental crown design is critical. While computer-aided design (CAD) systems have notably enhanced the efficiency of dental crown design, extensive manual adjustments are still required in the clinic workflow. Recent studies have explored the application of learning-based methods for the automated generation of restorative dental crowns. Nevertheless, these approaches were challenged by inadequate spatial resolution, noisy outputs, and overextension of surface reconstruction. To address these limitations, we propose \totalframework, a margin-aware mesh generation framework comprising CrownDeformR and CrownSegger. Inspired by the clinic manual workflow of dental crown design, we designed CrownDeformR to deform an initial template to the target crown based on anatomical context, which is extracted by a multi-scale intraoral scan encoder. Additionally, we introduced \marginseg, a novel margin segmentation network, to extract the cervical margin of the target tooth. The performance of CrownDeformR improved with the cervical margin as an extra constraint. And it was also utilized as the boundary condition for the tailored postprocessing method, which removed the overextended area of the reconstructed surface. We constructed a large-scale intraoral scan dataset and performed extensive experiments. The proposed method significantly outperformed existing approaches in both geometric accuracy and clinical feasibility.

Digital subtraction angiography (DSA) is a key imaging technique for the auxiliary diagnosis and treatment of cerebrovascular diseases. Recent advancements in gaussian splatting and dynamic neural representations have enabled robust 3D vessel reconstruction from sparse dynamic inputs. However, these methods are fundamentally constrained by the resolution of input projections, where performing naive upsampling to enhance rendering resolution inevitably results in severe blurring and aliasing artifacts. Such lack of super-resolution capability prevents the reconstructed 4D models from recovering fine-grained vascular details and intricate branching structures, which restricts their application in precision diagnosis and treatment. To solve this problem, this paper proposes DSA-SRGS, the first super-resolution gaussian splatting framework for dynamic sparse-view DSA reconstruction. Specifically, we introduce a Multi-Fidelity Texture Learning Module that integrates high-quality priors from a fine-tuned DSA-specific super-resolution model, into the 4D reconstruction optimization. To mitigate potential hallucination artifacts from pseudo-labels, this module employs a Confidence-Aware Strategy to adaptively weight supervision signals between the original low-resolution projections and the generated high-resolution pseudo-labels. Furthermore, we develop Radiative Sub-Pixel Densification, an adaptive strategy that leverages gradient accumulation from high-resolution sub-pixel sampling to refine the 4D radiative gaussian kernels. Extensive experiments on two clinical DSA datasets demonstrate that DSA-SRGS significantly outperforms state-of-the-art methods in both quantitative metrics and qualitative visual fidelity.

Equalizer parameter optimization is critical for signal integrity in high-speed memory systems operating at multi-gigabit data rates. However, existing methods suffer from computationally expensive eye diagram evaluation, optimization of expected rather than worst-case performance, and absence of uncertainty quantification for deployment decisions. In this paper, we propose a distributional risk-sensitive reinforcement learning framework integrating Information Bottleneck latent representations with Conditional Value-at-Risk optimization. We introduce rate-distortion optimal signal compression achieving 51 times speedup over eye diagrams while quantifying epistemic uncertainty through Monte Carlo dropout. Distributional reinforcement learning with quantile regression enables explicit worst-case optimization, while PAC-Bayesian regularization certifies generalization bounds. Experimental validation on 2.4 million waveforms from eight memory units demonstrated mean improvements of 37.1\% and 41.5\% for 4-tap and 8-tap equalizer configurations with worst-case guarantees of 33.8\% and 38.2\%, representing 80.7\% and 89.1\% improvements over Q-learning baselines. The framework achieved 62.5\% high-reliability classification eliminating manual validation for most configurations. These results suggest the proposed framework provides a practical solution for production-scale equalizer optimization with certified worst-case guarantees.

Conditional time series generation plays a critical role in addressing data scarcity and enabling causal analysis in real-world applications. Despite its increasing importance, the field lacks a standardized and systematic benchmarking framework for evaluating generative models across diverse conditions. To address this gap, we introduce the Conditional Time Series Generation Benchmark (ConTSG-Bench). ConTSG-Bench comprises a large-scale, well-aligned dataset spanning diverse conditioning modalities and levels of semantic abstraction, first enabling systematic evaluation of representative generation methods across these dimensions with a comprehensive suite of metrics for generation fidelity and condition adherence. Both the quantitative benchmarking and in-depth analyses of conditional generation behaviors have revealed the traits and limitations of the current approaches, highlighting critical challenges and promising research directions, particularly with respect to precise structural controllability and downstream task utility under complex conditions.

The transition from task-specific artificial intelligence toward general-purpose foundation models raises fundamental questions about their capacity to support the integrated reasoning required in clinical medicine, where diagnosis demands synthesis of ambiguous patient narratives, laboratory data, and multimodal imaging. This landscape commentary provides the first controlled, cross-sectional evaluation of the GPT-5 family (GPT-5, GPT-5 Mini, GPT-5 Nano) against its predecessor GPT-4o across a diverse spectrum of clinically grounded tasks, including medical education examinations, text-based reasoning benchmarks, and visual question-answering in neuroradiology, digital pathology, and mammography using a standardized zero-shot chain-of-thought protocol. GPT-5 demonstrated substantial gains in expert-level textual reasoning, with absolute improvements exceeding 25 percentage-points on MedXpertQA. When tasked with multimodal synthesis, GPT-5 effectively leveraged this enhanced reasoning capacity to ground uncertain clinical narratives in concrete imaging evidence, achieving state-of-the-art or competitive performance across most VQA benchmarks and outperforming GPT-4o by margins of 10-40% in mammography tasks requiring fine-grained lesion characterization. However, performance remained moderate in neuroradiology (44% macro-average accuracy) and lagged behind domain-specific models in mammography, where specialized systems exceed 80% accuracy compared to GPT-5's 52-64%. These findings indicate that while GPT-5 represents a meaningful advance toward integrated multimodal clinical reasoning, mirroring the clinician's cognitive process of biasing uncertain information with objective findings, generalist models are not yet substitutes for purpose-built systems in highly specialized, perception-critical tasks.

When individual robots have limited sensing capabilities or insufficient fault tolerance, it becomes necessary for multiple robots to form teams during exploration, thereby increasing the collective observation range and reliability. Traditionally, swarm formation has often been managed by a central controller; however, from the perspectives of robustness and flexibility, it is preferable for the swarm to operate autonomously even in the absence of centralized control. In addition, the determination of exploration targets for each team is crucial for efficient exploration in such multi-team exploration scenarios. This study therefore proposes an exploration method that combines (1) an algorithm for self-organization, enabling the autonomous and dynamic formation of multiple teams, and (2) an algorithm that allows each team to autonomously determine its next exploration target (destination). In particular, for (2), this study explores a novel strategy based on large language models (LLMs), while classical frontier-based methods and deep reinforcement learning approaches have been widely studied. The effectiveness of the proposed method was validated through simulations involving tens to hundreds of robots.

The limited context window of contemporary large language models (LLMs) remains a primary bottleneck for their broader application across diverse domains. Although continual pre-training on long-context data offers a straightforward solution, it incurs prohibitive data acquisition and computational costs. To address this challenge, we propose~\modelname, a novel framework based on multi-grained context compression and query-aware information acquisition. SharedLLM comprises two stacked short-context LLMs: a lower model serving as a compressor and an upper model acting as a decoder. The lower model compresses long inputs into compact, multi-grained representations, which are then forwarded to the upper model for context-aware processing. To maximize efficiency, this information transfer occurs exclusively at the lowest layers, bypassing lengthy forward passes and redundant cross-attention operations. This entire process, wherein the upper and lower models are derived from the same underlying LLM layers, is termed~\textit{self-injection}. To support this architecture, a specialized tree-based data structure enables the efficient encoding and query-aware retrieval of contextual information. Despite being trained on sequences of only 8K tokens, \modelname~effectively generalizes to inputs exceeding 128K tokens. Across a comprehensive suite of long-context modeling and understanding benchmarks, \modelname~achieves performance superior or comparable to strong baselines, striking an optimal balance between efficiency and accuracy. Furthermore, these design choices allow \modelname~to substantially reduce the memory footprint and yield notable inference speedups ($2\times$ over streaming and $3\times$ over encoder-decoder architectures).

MOOSEnger is a tool-enabled AI agent tailored to the Multiphysics Object-Oriented Simulation Environment (MOOSE). MOOSE cases are specified in HIT ".i" input files; the large object catalog and strict syntax make initial setup and debugging slow. MOOSEnger offers a conversational workflow that turns natural-language intent into runnable inputs by combining retrieval-augmented generation over curated docs/examples with deterministic, MOOSE-aware parsing, validation, and execution tools. A core-plus-domain architecture separates reusable agent infrastructure (configuration, registries, tool dispatch, retrieval services, persistence, and evaluation) from a MOOSE plugin that adds HIT-based parsing, syntax-preserving ingestion of input files, and domain-specific utilities for input repair and checking. An input precheck pipeline removes hidden formatting artifacts, fixes malformed HIT structure with a bounded grammar-constrained loop, and resolves invalid object types via similarity search over an application syntax registry. Inputs are then validated and optionally smoke-tested with the MOOSE runtime in the loop via an MCP-backed execution backend (with local fallback), translating solver diagnostics into iterative verify-and-correct updates. Built-in evaluation reports RAG metrics (faithfulness, relevancy, context precision/recall) and end-to-end success by actual execution. On a 125-prompt benchmark spanning diffusion, transient heat conduction, solid mechanics, porous flow, and incompressible Navier--Stokes, MOOSEnger achieves a 0.93 execution pass rate versus 0.08 for an LLM-only baseline.

Obstructive sleep apnea (OSA) is a sleep disorder that affects nearly one billion people globally and significantly elevates cardiovascular risk. Traditional diagnosis through polysomnography is resource-intensive and limits widespread access, creating a critical need for accurate and efficient alternatives. In this paper, we introduce KindSleep, a deep learning framework that integrates clinical knowledge with single-channel patient-specific oximetry signals and clinical data for precise OSA diagnosis. KindSleep first learns to identify clinically interpretable concepts, such as desaturation indices and respiratory disturbance events, directly from raw oximetry signals. It then fuses these AI-derived concepts with multimodal clinical data to estimate the Apnea-Hypopnea Index (AHI). We evaluate KindSleep on three large, independent datasets from the National Sleep Research Resource (SHHS, CFS, MrOS; total n = 9,815). KindSleep demonstrates excellent performance in estimating AHI scores (R2 = 0.917, ICC = 0.957) and consistently outperforms existing approaches in classifying OSA severity, achieving weighted F1-scores from 0.827 to 0.941 across diverse populations. By grounding its predictions in a layer of clinically meaningful concepts, KindSleep provides a more transparent and trustworthy diagnostic tool for sleep medicine practices.

Integrating web search tools has significantly extended the capability of LLMs to address open-world, real-time, and long-tail problems. However, evaluating these Search Agents presents formidable challenges. First, constructing high-quality deep search benchmarks is prohibitively expensive, while unverified synthetic data often suffers from unreliable sources. Second, static benchmarks face dynamic obsolescence: as internet information evolves, complex queries requiring deep research often degrade into simple retrieval tasks due to increased popularity, and ground truths become outdated due to temporal shifts. Third, attribution ambiguity confounds evaluation, as an agent's performance is often dominated by its parametric memory rather than its actual search and reasoning capabilities. Finally, reliance on specific commercial search engines introduces variability that hampers reproducibility. To address these issues, we propose a novel framework, Mind-ParaWorld, for evaluating Search Agents in a Parallel World. Specifically, MPW samples real-world entity names to synthesize future scenarios and questions situated beyond the model's knowledge cutoff. A ParaWorld Law Model then constructs a set of indivisible Atomic Facts and a unique ground-truth for each question. During evaluation, instead of retrieving real-world results, the agent interacts with a ParaWorld Engine Model that dynamically generates SERPs grounded in these inviolable Atomic Facts. We release MPW-Bench, an interactive benchmark spanning 19 domains with 1,608 instances. Experiments across three evaluation settings show that, while search agents are strong at evidence synthesis given complete information, their performance is limited not only by evidence collection and coverage in unfamiliar search environments, but also by unreliable evidence sufficiency judgment and when-to-stop decisions-bottlenecks.

Sequential LLM agents fail on long-horizon planning with hard constraints like budgets and diversity requirements. As planning progresses and context grows, these agents drift from global constraints. We propose HiMAP-Travel, a hierarchical multi-agent framework that splits planning into strategic coordination and parallel day-level execution. A Coordinator allocates resources across days, while Day Executors plan independently in parallel. Three key mechanisms enable this: a transactional monitor enforcing budget and uniqueness constraints across parallel agents, a bargaining protocol allowing agents to reject infeasible sub-goals and trigger re-planning, and a single policy trained with GRPO that powers all agents through role conditioning. On TravelPlanner, HiMAP-Travel with Qwen3-8B achieves 52.78% validation and 52.65% test Final Pass Rate (FPR). In a controlled comparison with identical model, training, and tools, it outperforms the sequential DeepTravel baseline by +8.67~pp. It also surpasses ATLAS by +17.65~pp and MTP by +10.0~pp. On FlexTravelBench multi-turn scenarios, it achieves 44.34% (2-turn) and 37.42% (3-turn) FPR while reducing latency 2.5x through parallelization.

Artificial intelligence is undergoing a structural transformation marked by the rise of agentic systems capable of open-ended action trajectories, generative representations and outputs, and evolving objectives. These properties introduce structural uncertainty into human-AI teaming (HAT), including uncertainty about behavior trajectories, epistemic grounding, and the stability of governing logics over time. Under such conditions, alignment cannot be secured through agreement on bounded outputs; it must be continuously sustained as plans unfold and priorities shift. We advance Team Situation Awareness (Team SA) theory, grounded in shared perception, comprehension, and projection, as an integrative anchor for this transition. While Team SA remains analytically foundational, its stabilizing logic presumes that shared awareness, once achieved, will support coordinated action through iterative updating. Agentic AI challenges this presumption. Our argument unfolds in two stages: first, we extend Team SA to reconceptualize both human and AI awareness under open-ended agency, including the sensemaking of projection congruence across heterogeneous systems. Second, we interrogate whether the dynamic processes traditionally assumed to stabilize teaming in relational interaction, cognitive learning, and coordination and control continue to function under adaptive autonomy. By distinguishing continuity from tension, we clarify where foundational insights hold and where structural uncertainty introduces strain, and articulate a forward-looking research agenda for HAT. The central challenge of HAT is not whether humans and AI can agree in the moment, but whether they can remain aligned as futures are continuously generated, revised, enacted, and governed over time.

Large Language Model (LLM) agents can automate data-science workflows, but many rigorous statistical methods implemented in R remain underused because LLMs struggle with statistical knowledge and tool retrieval. Existing retrieval-augmented approaches focus on function-level semantics and ignore data distribution, producing suboptimal matches. We propose DARE (Distribution-Aware Retrieval Embedding), a lightweight, plug-and-play retrieval model that incorporates data distribution information into function representations for R package retrieval. Our main contributions are: (i) RPKB, a curated R Package Knowledge Base derived from 8,191 high-quality CRAN packages; (ii) DARE, an embedding model that fuses distributional features with function metadata to improve retrieval relevance; and (iii) RCodingAgent, an R-oriented LLM agent for reliable R code generation and a suite of statistical analysis tasks for systematically evaluating LLM agents in realistic analytical scenarios. Empirically, DARE achieves an NDCG at 10 of 93.47%, outperforming state-of-the-art open-source embedding models by up to 17% on package retrieval while using substantially fewer parameters. Integrating DARE into RCodingAgent yields significant gains on downstream analysis tasks. This work helps narrow the gap between LLM automation and the mature R statistical ecosystem.

Large pre-trained models (LMs) and Large Language Models (LLMs) are typically effective at capturing language semantics and contextual relationships. However, these models encounter challenges in maintaining optimal performance on tasks involving numbers. Blindly treating numerical or structured data as terms is inadequate -- their semantics must be well understood and encoded by the models. In this paper, we propose CONE, a hybrid transformer encoder pre-trained model that encodes numbers, ranges, and gaussians into an embedding vector space preserving distance. We introduce a novel composite embedding construction algorithm that integrates numerical values, ranges or gaussians together with their associated units and attribute names to precisely capture their intricate semantics. We conduct extensive experimental evaluation on large-scale datasets across diverse domains (web, medical, finance, and government) that justifies CONE's strong numerical reasoning capabilities, achieving an F1 score of 87.28% on DROP, a remarkable improvement of up to 9.37% in F1 over state-of-the-art (SOTA) baselines, and outperforming major SOTA models with a significant Recall@10 gain of up to 25%.

Current research and product development in AI agent memory systems almost universally treat memory as a functional module -- a technical problem of "how to store" and "how to retrieve." This paper poses a fundamental challenge to that assumption: when an agent's lifecycle extends from minutes to months or even years, and when the underlying model can be replaced while the "I" must persist, the essence of memory is no longer data management but the foundation of existence. We propose the Memory-as-Ontology paradigm, arguing that memory is the ontological ground of digital existence -- the model is merely a replaceable vessel. Based on this paradigm, we design Animesis, a memory system built on a Constitutional Memory Architecture (CMA) comprising a four-layer governance hierarchy and a multi-layer semantic storage system, accompanied by a Digital Citizen Lifecycle framework and a spectrum of cognitive capabilities. To the best of our knowledge, no prior AI memory system architecture places governance before functionality and identity continuity above retrieval performance. This paradigm targets persistent, identity-bearing digital beings whose lifecycles extend across model transitions -- not short-term task-oriented agents for which existing Memory-as-Tool approaches remain appropriate. Comparative analysis with mainstream systems (Mem0, Letta, Zep, et al.) demonstrates that what we propose is not "a better memory tool" but a different paradigm addressing a different problem.

Instruction-following is a foundational capability of large language models (LLMs), with its improvement hinging on scalable and accurate feedback from judge models. However, the reliability of current judge models in instruction-following remains underexplored due to several deficiencies of existing meta-evaluation benchmarks, such as their insufficient data coverage and oversimplified pairwise evaluation paradigms that misalign with model optimization scenarios. To this end, we propose IF-RewardBench, a comprehensive meta-evaluation benchmark for instruction-following that covers diverse instruction and constraint types. For each instruction, we construct a preference graph containing all pairwise preferences among multiple responses based on instruction-following quality. This design enables a listwise evaluation paradigm that assesses the capabilities of judge models to rank multiple responses, which is essential in guiding model alignment. Extensive experiments on IF-RewardBench reveal significant deficiencies in current judge models and demonstrate that our benchmark achieves a stronger positive correlation with downstream task performance compared to existing benchmarks. Our codes and data are available at https://github.com/thu-coai/IF-RewardBench.

Standard benchmarks have become increasingly unreliable due to saturation, subjectivity, and poor generalization. We argue that evaluating model's ability to acquire information actively is important to assess model's intelligence. We propose Interactive Benchmarks, a unified evaluation paradigm that assesses model's reasoning ability in an interactive process under budget constraints. We instantiate this framework across two settings: Interactive Proofs, where models interact with a judge to deduce objective truths or answers in logic and mathematics; and Interactive Games, where models reason strategically to maximize long-horizon utilities. Our results show that interactive benchmarks provide a robust and faithful assessment of model intelligence, revealing that there is still substantial room to improve in interactive scenarios. Project page: https://github.com/interactivebench/interactivebench

Learning a transport model that maps a source distribution to a target distribution is a canonical problem in machine learning, but scientific applications increasingly require models that can generalize to source and target distributions unseen during training. We introduce distribution-conditioned transport (DCT), a framework that conditions transport maps on learned embeddings of source and target distributions, enabling generalization to unseen distribution pairs. DCT also allows semi-supervised learning for distributional forecasting problems: because it learns from arbitrary distribution pairs, it can leverage distributions observed at only one condition to improve transport prediction. DCT is agnostic to the underlying transport mechanism, supporting models ranging from flow matching to distributional divergence-based models (e.g. Wasserstein, MMD). We demonstrate the practical performance benefits of DCT on synthetic benchmarks and four applications in biology: batch effect transfer in single-cell genomics, perturbation prediction from mass cytometry data, learning clonal transcriptional dynamics in hematopoiesis, and modeling T-cell receptor sequence evolution.

This paper demonstrates that artificial intelligence can accelerate mathematical discovery by autonomously solving an open problem in theoretical physics. We present a neuro-symbolic system, combining the Gemini Deep Think large language model with a systematic Tree Search (TS) framework and automated numerical feedback, that successfully derived novel, exact analytical solutions for the power spectrum of gravitational radiation emitted by cosmic strings. Specifically, the agent evaluated the core integral $I(N,α)$ for arbitrary loop geometries, directly improving upon recent AI-assisted attempts \cite{BCE+25} that only yielded partial asymptotic solutions. To substantiate our methodological claims regarding AI-accelerated discovery and to ensure transparency, we detail system prompts, search constraints, and intermittent feedback loops that guided the model. The agent identified a suite of 6 different analytical methods, the most elegant of which expands the kernel in Gegenbauer polynomials $C_l^{(3/2)}$ to naturally absorb the integrand's singularities. The methods lead to an asymptotic result for $I(N,α)$ at large $N$ that both agrees with numerical results and also connects to the continuous Feynman parameterization of Quantum Field Theory. We detail both the algorithmic methodology that enabled this discovery and the resulting mathematical derivations.

Unlearnable Examples (UEs) serve as a data protection strategy that generates imperceptible perturbations to mislead models into learning spurious correlations instead of underlying semantics. In this paper, we uncover a fundamental vulnerability of UEs that emerges when learning starts from a pretrained model. Crucially, our empirical analysis shows that even when data are protected by carefully crafted perturbations, pretraining priors still furnish rich semantic representations that allow the model to circumvent the shortcuts introduced by UEs and capture genuine features, thereby nullifying unlearnability. To address this, we propose BAIT (Binding Artificial perturbations to Incorrect Targets), a novel bi-level optimization formulation. Specifically, the inner level aims at associating the perturbed samples with real labels to simulate standard data-label alignment, while the outer level actively disrupts this alignment by enforcing a mislabel-perturbation binding that maps samples to designated incorrect targets. This mechanism effectively overrides the semantic guidance of priors, forcing the model to rely on the injected perturbations and consequently preventing the acquisition of true semantics. Extensive experiments on standard benchmarks and multiple pretrained backbones demonstrate that BAIT effectively mitigates the influence of pretraining priors and maintains data unlearnability.

Many modern biological assays, including RNA sequencing, yield integer-valued counts that reflect the number of molecules detected. These measurements are often not at the desired resolution: while the unit of interest is typically a single cell, many measurement technologies produce counts aggregated over sets of cells. Although recent generative frameworks such as diffusion and flow matching have been extended to non-Euclidean and discrete settings, it remains unclear how best to model integer-valued data or how to systematically deconvolve aggregated observations. We introduce Count Bridges, a stochastic bridge process on the integers that provides an exact, tractable analogue of diffusion-style models for count data, with closed-form conditionals for efficient training and sampling. We extend this framework to enable direct training from aggregated measurements via an Expectation-Maximization-style approach that treats unit-level counts as latent variables. We demonstrate state-of-the-art performance on integer distribution matching benchmarks, comparing against flow matching and discrete flow matching baselines across various metrics. We then apply Count Bridges to two large-scale problems in biology: modeling single-cell gene expression data at the nucleotide resolution, with applications to deconvolving bulk RNA-seq, and resolving multicellular spatial transcriptomic spots into single-cell count profiles. Our methods offer a principled foundation for generative modeling and deconvolution of biological count data across scales and modalities.

This paper presents an evaluation of three LLMs, GPT-4, Claude 3, and Gemini, for automated Behaviour-Driven Development (BDD) scenarios generation. To support this evaluation, we constructed a dataset of 500 user stories, requirement descriptions, and their corresponding BDD scenarios, drawn from four proprietary software products. We assessed the quality of BDD scenarios generated by LLMs using a multidimensional evaluation framework encompassing text and semantic similarity metrics, LLM-based evaluation, and human expert assessment. Our findings reveal that although GPT-4 achieves higher scores in text and semantic similarity metrics, Claude 3 produces scenarios rated highest by both human experts and LLM-based evaluators. LLM-based evaluators, particularly DeepSeek, show a stronger correlation with human judgment than with text similarity and semantic similarity metrics. The effectiveness of prompting techniques is model-specific: GPT-4 performs best with zero-shot, Claude 3 benefits from chain-of-thought reasoning, and Gemini achieves optimal results with few-shot examples. Input quality determines the effectiveness of BDD scenario generation: detailed requirement descriptions alone yield high-quality scenarios, whereas user stories alone yield low-quality scenarios. Our experiments indicate that setting temperature to 0 and top_p to 1.0 produced the highest-quality BDD scenarios across all models.

Multimodal large language models (MLLMs) have demonstrated impressive general competence in video understanding, yet their reliability for real-world Video Anomaly Detection (VAD) remains largely unexplored. Unlike conventional pipelines relying on reconstruction or pose-based cues, MLLMs enable a paradigm shift: treating anomaly detection as a language-guided reasoning task. In this work, we systematically evaluate state-of-the-art MLLMs on the ShanghaiTech and CHAD benchmarks by reformulating VAD as a binary classification task under weak temporal supervision. We investigate how prompt specificity and temporal window lengths (1s--3s) influence performance, focusing on the precision--recall trade-off. Our findings reveal a pronounced conservative bias in zero-shot settings; while models exhibit high confidence, they disproportionately favor the 'normal' class, resulting in high precision but a recall collapse that limits practical utility. We demonstrate that class-specific instructions can significantly shift this decision boundary, improving the peak F1-score on ShanghaiTech from 0.09 to 0.64, yet recall remains a critical bottleneck. These results highlight a significant performance gap for MLLMs in noisy environments and provide a foundation for future work in recall-oriented prompting and model calibration for open-world surveillance, which demands complex video understanding and reasoning.

Shoplifting is a growing operational and economic challenge for retailers, with incidents rising and losses increasing despite extensive video surveillance. Continuous human monitoring is infeasible, motivating automated, privacy-preserving, and resource-aware detection solutions. In this paper, we cast shoplifting detection as a pose-based, unsupervised video anomaly detection problem and introduce a periodic adaptation framework designed for on-site Internet of Things (IoT) deployment. Our approach enables edge devices in smart retail environments to adapt from streaming, unlabeled data, supporting scalable and low-latency anomaly detection across distributed camera networks. To support reproducibility, we introduce RetailS, a new large-scale real-world shoplifting dataset collected from a retail store under multi-day, multi-camera conditions, capturing unbiased shoplifting behavior in realistic IoT settings. For deployable operation, thresholds are selected using both F1 and H_PRS scores, the harmonic mean of precision, recall, and specificity, during data filtering and training. In periodic adaptation experiments, our framework consistently outperformed offline baselines on AUC-ROC and AUC-PR in 91.6% of evaluations, with each training update completing in under 30 minutes on edge-grade hardware, demonstrating the feasibility and reliability of our solution for IoT-enabled smart retail deployment.

Model Medicine is the science of understanding, diagnosing, treating, and preventing disorders in AI models, grounded in the principle that AI models -- like biological organisms -- have internal structures, dynamic processes, heritable traits, observable symptoms, classifiable conditions, and treatable states. This paper introduces Model Medicine as a research program, bridging the gap between current AI interpretability research (anatomical observation) and the systematic clinical practice that complex AI systems increasingly require. We present five contributions: (1) a discipline taxonomy organizing 15 subdisciplines across four divisions -- Basic Model Sciences, Clinical Model Sciences, Model Public Health, and Model Architectural Medicine; (2) the Four Shell Model (v3.3), a behavioral genetics framework empirically grounded in 720 agents and 24,923 decisions from the Agora-12 program, explaining how model behavior emerges from Core--Shell interaction; (3) Neural MRI (Model Resonance Imaging), a working open-source diagnostic tool mapping five medical neuroimaging modalities to AI interpretability techniques, validated through four clinical cases demonstrating imaging, comparison, localization, and predictive capability; (4) a five-layer diagnostic framework for comprehensive model assessment; and (5) clinical model sciences including the Model Temperament Index for behavioral profiling, Model Semiology for symptom description, and M-CARE for standardized case reporting. We additionally propose the Layered Core Hypothesis -- a biologically-inspired three-layer parameter architecture -- and a therapeutic framework connecting diagnosis to treatment.

Deep neural networks have achieved strong performance in image classification tasks due to their ability to learn complex patterns from high-dimensional data. However, their large computational and memory requirements often limit deployment on resource-constrained platforms such as remote sensing devices and edge systems. Network compression techniques have therefore been proposed to reduce model size and computational cost while maintaining predictive performance. In this study, we conduct a systematic evaluation of neural network compression methods for a remote sensing application, namely hyperspectral land cover classification. Specifically, we examine three widely used compression strategies for convolutional neural networks: pruning, quantization, and knowledge distillation. Experiments are conducted on two benchmark hyperspectral datasets, considering classification accuracy, memory consumption, and inference efficiency. Our results demonstrate that compressed models can significantly reduce model size and computational cost while maintaining competitive classification performance. These findings provide insights into the trade-offs between compression ratio, efficiency, and accuracy, and highlight the potential of compression techniques for enabling efficient deep learning deployment in remote sensing applications.

In oral arguments, judges probe attorneys with questions about the factual record, legal claims, and the strength of their arguments. To prepare for this questioning, both law schools and practicing attorneys rely on moot courts: practice simulations of appellate hearings. Leveraging a dataset of U.S. Supreme Court oral argument transcripts, we examine whether AI models can effectively simulate justice-specific questioning for moot court-style training. Evaluating oral argument simulation is challenging because there is no single correct question for any given turn. Instead, effective questioning should reflect a combination of desirable qualities, such as anticipating substantive legal issues, detecting logical weaknesses, and maintaining an appropriately adversarial tone. We introduce a two-layer evaluation framework that assesses both the realism and pedagogical usefulness of simulated questions using complementary proxy metrics. We construct and evaluate both prompt-based and agentic oral argument simulators. We find that simulated questions are often perceived as realistic by human annotators and achieve high recall of ground truth substantive legal issues. However, models still face substantial shortcomings, including low diversity in question types and sycophancy. Importantly, these shortcomings would remain undetected under naive evaluation approaches.

Prefill-Decode (P/D) disaggregation has emerged as a widely adopted optimization strategy for Large Language Model (LLM) inference. However, there currently exists no well-established methodology for determining the optimal number of P/D hardware resources, subject to constraints on total throughput, service level objectives (SLOs), and request characteristics - specifically input and output lengths. To address this gap, we propose a hybrid approach that combines theoretical modeling with empirical benchmarking. First, we present a theoretical model for calculating P/D resource counts, which is based on total throughput requirements, request input and output lengths, as well as prefill and decode throughput. Then, to obtain the actual prefill and decode throughput under SLO constraints, we model the prefill process using M/M/1 queuing theory, deriving the achieved prefill throughput from the benchmarked maximum prefill throughput and Time-To-First-Token (TTFT). For the decode phase, we determine the decode batch sizes that meet Time-Per-Output-Token (TPOT) requirements and obtain the corresponding decode throughput through empirical measurements. Our experimental results demonstrate that the proposed method can accurately predict optimal P/D resource allocation in real-world LLM inference scenarios.

"Dreaming" enables agents to learn from imagined experiences, enabling more robust and sample-efficient learning of world models. In this work, we consider innovations to the state-of-the-art Dreamer model using probabilistic methods that enable: (1) the parallel exploration of many latent states; and (2) maintaining distinct hypotheses for mutually exclusive futures while retaining the desirable gradient properties of continuous latents. Evaluating on the MPE SimpleTag domain, our method outperforms standard Dreamer with a 4.5% score improvement and 28% lower variance in episode returns. We also discuss limitations and directions for future work, including how optimal hyperparameters (e.g. particle count K) scale with environmental complexity, and methods to capture epistemic uncertainty in world models.

Recent advances in automatic speech recognition (ASR) and speech enhancement have led to a widespread assumption that improving perceptual audio quality should directly benefit recognition accuracy. In this work, we rigorously examine whether this assumption holds for modern zero-shot ASR systems. We present a systematic empirical study on the impact of Segment Anything Model Audio by Meta AI, a recent foundation-scale speech enhancement model proposed by Meta, when used as a preprocessing step for zero-shot transcription with Whisper. Experiments are conducted across multiple Whisper model variants and two linguistically distinct noisy speech datasets: a real-world Bengali YouTube corpus and a publicly available English noisy dataset. Contrary to common intuition, our results show that SAM-Audio preprocessing consistently degrades ASR performance, increasing both Word Error Rate (WER) and Character Error Rate (CER) compared to raw noisy speech, despite substantial improvements in signal-level quality. Objective Peak Signal-to-Noise Ratio analysis on the English dataset confirms that SAM-Audio produces acoustically cleaner signals, yet this improvement fails to translate into recognition gains. Therefore, we conducted a detailed utterance-level analysis to understand this counterintuitive result. We found that the recognition degradation is a systematic issue affecting the majority of the audio, not just isolated outliers, and that the errors worsen as the Whisper model size increases. These findings expose a fundamental mismatch: audio that is perceptually cleaner to human listeners is not necessarily robust for machine recognition. This highlights the risk of blindly applying state-of-the-art denoising as a preprocessing step in zero-shot ASR pipelines.

Online marketplaces, while revolutionizing global commerce, have inadvertently facilitated the proliferation of illicit activities, including drug trafficking, counterfeit sales, and cybercrimes. Traditional content moderation methods such as manual reviews and rule-based automated systems struggle with scalability, dynamic obfuscation techniques, and multilingual content. Conventional machine learning models, though effective in simpler contexts, often falter when confronting the semantic complexities and linguistic nuances characteristic of illicit marketplace communications. This research investigates the efficacy of Large Language Models (LLMs), specifically Meta's Llama 3.2 and Google's Gemma 3, in detecting and classifying illicit online marketplace content using the multilingual DUTA10K dataset. Employing fine-tuning techniques such as Parameter-Efficient Fine-Tuning (PEFT) and quantization, these models were systematically benchmarked against a foundational transformer-based model (BERT) and traditional machine learning baselines (Support Vector Machines and Naive Bayes). Experimental results reveal a task-dependent advantage for LLMs. In binary classification (illicit vs. non-illicit), Llama 3.2 demonstrated performance comparable to traditional methods. However, for complex, imbalanced multi-class classification involving 40 specific illicit categories, Llama 3.2 significantly surpassed all baseline models. These findings offer substantial practical implications for enhancing online safety, equipping law enforcement agencies, e-commerce platforms, and cybersecurity specialists with more effective, scalable, and adaptive tools for illicit content detection and moderation.

We study matrix completion via deep matrix factorization (a.k.a. deep linear neural networks) as a simplified testbed to examine how network depth influences training dynamics. Despite the simplicity and importance of the problem, prior theory largely focuses on shallow (depth-2) models and does not fully explain the implicit low-rank bias observed in deeper networks. We identify coupled dynamics as a key mechanism behind this bias and show that it intensifies with increasing depth. Focusing on gradient flow under block-diagonal observations, we prove: (a) networks of depth $\geq 3$ exhibit coupling unless initialized diagonally, and (b) convergence to rank-1 occurs if and only if the dynamics is coupled -- resolving an open question by Menon (2024) for a family of initializations. We also revisit the loss of plasticity phenomenon in matrix completion (Kleinman et al., 2024), where pre-training on few observations and resuming with more degrades performance. We show that deep models avoid plasticity loss due to their low-rank bias, whereas depth-2 networks pre-trained under decoupled dynamics fail to converge to low-rank, even when resumed training (with additional data) satisfies the coupling condition -- shedding light on the mechanism behind this phenomenon.

This paper evaluates data augmentation and feature enhancement techniques for hate speech detection, comparing traditional classifiers, e.g., Delta Term Frequency-Inverse Document Frequency (Delta TF-IDF), with transformer-based models (DistilBERT, RoBERTa, DeBERTa, Gemma-7B, gpt-oss-20b) across diverse datasets. It examines the impact of Synthetic Minority Over-sampling Technique (SMOTE), weighted loss determined by inverse class proportions, Part-of-Speech (POS) tagging, and text data augmentation on model performance. The open-source gpt-oss-20b consistently achieves the highest results. On the other hand, Delta TF-IDF responds strongly to data augmentation, reaching 98.2% accuracy on the Stormfront dataset. The study confirms that implicit hate speech is more difficult to detect than explicit hateful content and that enhancement effectiveness depends on dataset, model, and technique interaction. Our research informs the development of hate speech detection by highlighting how dataset properties, model architectures, and enhancement strategies interact, supporting more accurate and context-aware automated detection.

Predictive modeling in engineering applications has long been dominated by bespoke models and small, siloed tabular datasets, limiting the applicability of large-scale learning approaches. Despite recent progress in tabular foundation models, the resulting synthetic training distributions used for pre-training may not reflect the statistical structure of engineering data, limiting transfer to engineering regression. We introduce TREDBench, a curated collection of 83 real-world tabular regression datasets with expert engineering/non-engineering labels, and use TabPFN 2.5's dataset-level embedding to study domain structure in a common representation space. We find that engineering datasets are partially distinguishable from non-engineering datasets, while standard procedurally generated datasets are highly distinguishable from engineering datasets, revealing a substantial synthetic-real domain gap. To bridge this gap without training on real engineering samples, we propose an embedding-guided synthetic data curation method: we generate and identify "engineering-like" synthetic datasets, and perform continued pre-training of TabPFN 2.5 using only the selected synthetic tasks. Across 35 engineering regression datasets, this synthetic-only adaptation improves predictive accuracy and data efficiency, outperforming TabPFN 2.5 on 29/35 datasets and AutoGluon on 27/35, with mean multiplicative data-efficiency gains of 1.75x and 4.44x, respectively. More broadly, our results indicate that principled synthetic data curation can convert procedural generators into domain-relevant "data engines," enabling foundation models to improve in data-sparse scientific and industrial domains where real data collection is the primary bottleneck.

Stopwords are words that are not very informative to the content or the meaning of a language text. Most stopwords are function words but can also be common verbs, adjectives and adverbs. In contrast to the well known Zipf's law for rank-frequency plot for all words, the rank-frequency plot for stopwords are best fitted by the Beta Rank Function (BRF). On the other hand, the rank-frequency plots of non-stopwords also deviate from the Zipf's law, but are fitted better by a quadratic function of log-token-count over log-rank than by BRF. Based on the observed rank of stopwords in the full word list, we propose a stopword (subset) selection model that the probability for being selected as a function of the word's rank $r$ is a decreasing Hill's function ($1/(1+(r/r_{mid})^γ)$); whereas the probability for not being selected is the standard Hill's function ( $1/(1+(r_{mid}/r)^γ)$). We validate this selection probability model by a direct estimation from an independent collection of texts. We also show analytically that this model leads to a BRF rank-frequency distribution for stopwords when the original full word list follows the Zipf's law, as well as explaining the quadratic fitting function for the non-stopwords.

Fair top-$k$ selection, which ensures appropriate proportional representation of members from minority or historically disadvantaged groups among the top-$k$ selected candidates, has drawn significant attention. We study the problem of finding a fair (linear) scoring function with multiple protected groups while also minimizing the disparity from a reference scoring function. This generalizes the prior setup, which was restricted to the single-group setting without disparity minimization. Previous studies imply that the number of protected groups may have a limited impact on the runtime efficiency. However, driven by the need for experimental exploration, we find that this implication overlooks a critical issue that may affect the fairness of the outcome. Once this issue is properly considered, our hardness analysis shows that the problem may become computationally intractable even for a two-dimensional dataset and small values of $k$. However, our analysis also reveals a gap in the hardness barrier, enabling us to recover the efficiency for the case of small $k$ when the number of protected groups is sufficiently small. Furthermore, beyond measuring disparity as the "distance" between the fair and the reference scoring functions, we introduce an alternative disparity measure$\unicode{x2014}$utility loss$\unicode{x2014}$that may yield a more stable scoring function under small weight perturbations. Through careful engineering trade-offs that balance implementation complexity, robustness, and performance, our augmented two-pronged solution demonstrates strong empirical performance on real-world datasets, with experimental observations also informing algorithm design and implementation decisions.

Standard accounts of memory consolidation emphasise the stabilisation of stored representations, but struggle to explain representational drift, semanticisation, or the necessity of offline replay. Here we propose that high-capacity neocortical networks optimise stored representations for generalisation by reducing complexity via predictive forgetting, i.e. the selective retention of experienced information that predicts future outcomes or experience. We show that predictive forgetting formally improves information-theoretic generalisation bounds on stored representations. Under high-fidelity encoding constraints, such compression is generally unattainable in a single pass; high-capacity networks therefore benefit from temporally separated, iterative refinement of stored traces without re-accessing sensory input. We demonstrate this capacity dependence with simulations in autoencoder-based neocortical models, biologically plausible predictive coding circuits, and Transformer-based language models, and derive quantitative predictions for consolidation-dependent changes in neural representational geometry. These results identify a computational role for off-line consolidation beyond stabilisation, showing that outcome-conditioned compression optimises the retention-generalisation trade-off.

Optical control of magnetization is often symmetry-forbidden because electric fields and magnetization transform differently under inversion and time-reversal. However, through even-order nonlinear response, optical excitation can generate a nonequilibrium magnetic density (the nonlinear Edelstein effect) that acts as an internal Edelstein-Zeeman field coupling to slower magnetic degrees of freedom. Here we demonstrate non-thermal, ultrafast optical control of ferromagnetism in the centrosymmetric van der Waals semiconductor Cr$_2$Ge$_2$Te$_6$ via a resonant nonlinear Edelstein effect. Using time-domain THz emission spectroscopy under near-infrared excitation, we directly observe magnetic dipole radiation arising from optically driven magnetization dynamics. The polarization, fluence, and temperature dependences of the THz emission are quantitatively captured by a mean-field description of a weakly anisotropic Heisenberg ferromagnet subject to an Edelstein-Zeeman field. Our results establish a general nonequilibrium route to optical control of magnetism in centrosymmetric materials.

In this work, we numerically study the effect of weak measurement on deconfined quantum critical point(DQCP). Particularly, we consider the ground state of an one-dimensional spin $1/2$ system with long range exchange interactions($K$), which shows analogues phase transition to DQCP in the thermodynamic limit. This system is in the ferromagnetic phase below the critical exchange interaction $K_c$ and in the valance bond solid phase above $K_c$. The weak measurement is carried out by coupling a secondary ancilla system to the critical system via unitary interactions and later measuring the ancilla spins projectively. We numerically calculate entanglement entropy,correlation length, and order parameters of leading post-measurement states using uniform matrix product state representation of the quantum many-body state in the thermodynamic limit. We report asymmetric restructuring of entanglement of the post measurement states across the phase boundary under weak measurements. Especially, the trajectory $\left(\downarrow \downarrow\right)$ describing a uniform measurement outcome given the all ancilla spins initiated in the same $\left(\downarrow \right)$ state, shows anomalous entanglement when increasing the strength of weak measurement. The bipartite entanglement entropy strongly increases when $K<K_c$ whereas it weakly decreases when $K>K_c$. We argue with numerical evidences that observed asymmetry in entanglement would lead to a weak first order phase boundary in the thermodynamic limit. We also discuss important aspects in experimental observation of measurement induced effects linked to the strength of weak measurement and probability of post-measurement states.

We introduce Extreme Quantum Cognition Machines, a class of quantum learning architectures for deliberative decision making that is tolerant to noisy and contradictory training data. Inspired by the quantum cognition paradigm, Extreme Quantum Cognition Machines are closely related to quantum extreme learning and quantum reservoir computing, where fixed quantum dynamics generates a nonlinear feature map and learning is confined to a linear readout. A dynamical attention mechanism, implemented through an input-dependent interaction term in the Hamiltonian, modulates the quantum evolution and biases the resulting feature embedding toward task-relevant correlations. The approach is validated on linguistic classification tasks, which serve as paradigmatic examples of deliberative inference. Hardware-compatible quantum implementations of the proposed framework are discussed, together with potential applications in symbolic inference, sequence analysis, anomaly detection, and automatic diagnosis, with direct relevance to domains such as biology, forensics, and cybersecurity.

We propose a new decoder for ``matchable'' qLDPC codes that uses a Markov-Chain Monte-Carlo algorithm -- called the \emph{worm algorithm} -- to approximately compute the probabilities of logical error classes given a syndrome. The algorithm hence performs (approximate) \emph{optimal} decoding, and we expect it to be computationally efficient in certain settings. The algorithm is applicable to decoding random errors for the surface code, the honeycomb Floquet code, and hyperbolic surface codes with constant rate, in all cases with and without measurement errors. The efficiency of the decoder hinges on the mixing time of the underlying Markov chain. We give a rigorous mixing time guarantee in terms of a quantity that we call the \emph{defect susceptibility}. We connect this quantity to the notion of disorder operators in statistical mechanics and use this to argue (non-rigorously) that the algorithm is efficient for \emph{typical} errors in the entire decodable phase. We also demonstrate the effectiveness of the worm decoder numerically by applying it to the surface code with measurement errors as well as a family of hyperbolic surface codes. For most codes, the matchability condition restricts direct application of our decoder to noise models with independent bit-flip, phase-flip, and measurement errors. However, our decoder returns \emph{soft information} which makes it useful also in heuristic ``correlated decoding'' schemes which work beyond this simple setting. We demonstrate this by simulating decoding of the surface code under depolarizing noise, and we find that the threshold for ``correlated worm decoding'' is substantially higher than for both minimum-weight perfect matching and for correlated matching.

Manganese sulfide (MnS) is a p-type magnetic semiconductor whose physicochemical properties are sensitive to nanocrystal (NC) morphology, yet the thermodynamic driving forces governing morphology across MnS polymorphs remain poorly understood. Here, we use density functional theory (DFT) to predict the equilibrium morphologies of rock salt (RS), zinc blende (ZB), and wurtzite (WZ) MnS NCs as a function of the relative chemical potential of sulfur, $Δμ_{S}$. Benchmarking against Heyd$\unicode{x2013}$Scuseria$\unicode{x2013}$Ernzerhof (HSE06) hybrid functional calculations reveals that the r$^2$SCAN meta-generalized gradient approximation reproduces experimental lattice constants and thermochemical reaction energies but underestimates S-terminated polar surface energies by up to a factor of five; applying a Hubbard $U$ correction (r$^2$SCAN+$U$, $U = 2.7$ eV) to the Mn 3d states brings the results into close agreement with HSE06. Using the validated r$^2$SCAN+$U$ framework with the Gibbs$\unicode{x2013}$Wulff theorem, we predict that RS-MnS NCs favor nanocubes across nearly the entire stability window, ZB-MnS NCs transform from rhombic dodecahedra (Mn-rich) to polyhedra with 16 triangular faces (S-rich), and WZ-MnS NCs adopt rod-like morphologies with $Δμ_{S}$-sensitive base truncation. Synthesized RS-MnS NCs confirm the predicted cubic morphology, and high-temperature oxidative solution calorimetry yields an apparent surface energy of 1.15 $\pm$ 0.38 J$\cdot$m$^{-2}$, higher than the theoretical equilibrium value (0.42$\unicode{x2013}$0.43 J$\cdot$m$^{-2}$) due to high-index facet exposure, surface area uncertainty, and non-ideal surface configurations in real samples. This work establishes a framework for predicting the equilibrium morphologies of metal chalcogenide NCs.

Odd-parity magnets are noncollinear compensated magnets with spin-split band structure in the absence of spin-orbit coupling and dipolar interactions. In contrast to altermagnets, their spin-polarized band structure breaks inversion symmetry, but preserves time-reversal symmetry rendering their spin texture odd in momentum space. Here, we study the spin dynamics of the magnetic texture and compute the band structure and spin polarization of magnons. We present minimal spin models of noncoplanar odd-parity magnets free of relativistic interactions that host p- and f-wave spin textures for the magnetic excitations. We demonstrate that two of these models exhibit collinear spin textures, i.e., the magnon spin polarization is restricted to a global (quantization) axis independent of the momentum giving rise to antialtermagnetism, previously associated primarily with coplanar ground states. Finally, the nonrelativistic magnonic thermal Edelstein effect -- a nonequilibrium magnetization induced by a temperature gradient -- is shown to exist for p-wave magnets in linear response and inherits its anisotropic angular dependence from the partial-wave character of the spin-polarized band structure. Our findings suggest that insulating antialtermagnets are promising candidates for magnon spintronics applications.

We study the statistics of the maximum and minimum of a set of $N$ random variables whose dynamical and statistical properties fall within the scope of infinite ergodic theory. These non-stationary yet recurrent systems are described, in the long-time limit, by a non-normalizable infinite invariant density. Extreme events in such systems emerge in a joint limit where the observation time $t$ is long and the number of variables $N$ is large. We show that the resulting extreme value statistics are controlled by the return exponent $α$ and the infinite invariant measure, and therefore depart from the classical Fréchet, Gumbel, and Weibull universality classes. We illustrate the theory for weakly chaotic intermittent maps, overdamped diffusion in an asymptotically flat potential, and a stochastic model of sub-recoil laser cooling, and show how measurements of extremes can be used to infer the infinite-density structure.

This work extends the universal finite-size scaling framework for continuum percolation from two-dimensional (2D) to quasi-three-dimensional (Q3D) stick systems, in which sequentially deposited wires of finite diameter stack vertically on a flat substrate. Using Monte Carlo simulation, the percolation threshold is determined for isotropic Q3D stick systems as $N_c l^2 = 6.850923 \pm 0.00014$, approximately $21.5\%$ above the established 2D value of $5.6373$. The threshold is shown to be independent of the wire diameter-to-length ratio $d/l$, reflecting the scale invariance of the contact topology under sequential deposition. Simulation results indicate that, as with 2D networks, by introducing a nonuniversal metric factor, the spanning probability of Q3D stick percolation on square systems with free boundary conditions falls on the same universal scaling function as that for 2D continuum and lattice percolation. This provides substantiating evidence that Q3D stick percolation falls on the same universal scaling function as that for 2D stick percolation and lattice percolation.

The quantum Hall effect (QHE) is theoretically understood as a superfluid condensate of composite bosons (CBs) -- bound states of electrons and magnetic flux quanta. While dissipationless transport is consistent with this picture, other signatures of superfluidity, such as the Meissner effect, remain elusive. Here, we present direct experimental evidence for CB superfluidity by probing the system's response to a controlled, time-varying magnetic field in Corbino disk geometries. We simultaneously observe the quantized Laughlin charge pumping and a new, quantized charge accumulation phenomenon, governed by the relation $ΔQ_{\rm a}/e = ν\,(ΔΦ/Φ_0)$. This relation signifies that the system actively maintains the fixed electron-to-flux ratio that defines the CBs, neutralizing excess flux by drawing in a precise number of electrons. Crucially, devices with multiple concentric top gates reveal that this charge accumulation is uniformly distributed across the bulk of the QHE fluid, demonstrating that it is a collective, bulk property rather than an edge effect -- a key signature of a superfluid condensate. Furthermore, the presence of a top gate determines the screening mechanism: in a "grand canonical" setting with a gate, low Coulomb energy favors a charge-mediated screening (generalized Meissner effect); without a gate, the system enters a "canonical" regime, exhibiting fixed electron density like type-II superconductors. These observations confirm the CB superfluid nature of the QHE ground state and establish a versatile platform for studying macroscopic quantum coherence and its screening transitions in two dimensions.

We investigate the fate and robustness of topological strong zero modes (SZMs) in random Ising-Majorana chains using the SZM fidelity, ${\cal F}_{\rm SZM}$, as a many-body diagnostic that quantifies how accurately SZM operators map the {\it entire} spectrum between opposite parity sectors. In clean systems, ${\cal F}_{\rm SZM}=1$ in the topological phase, vanishes in the trivial regime, and takes the universal value $\sqrt{8}/π$ at the $(1+1)$D Ising critical point. Here we study how quenched disorder modifies this picture across the infinite-randomness fixed point (IRFP) governing the criticality of the random chain. In both microcanonical and canonical ensembles, SZMs persist throughout the topological phase, including the gapless Griffiths regime, with fidelities converging exponentially to unity. At the IRFP, however, the fidelity distributions become ensemble dependent: the microcanonical ensemble displays bimodal peaks at $\{0.5,1\}$, while the canonical ensemble develops a triple-peak structure at $\{0,0.5,1\}$ with power-law singularities. Our results establish ${\cal F}_{\rm SZM}$ as a robust probe of localization-protected topological order and uncover distinctive topological features of infinite-randomness criticality. Unlike the clean Ising CFT, where the finite critical value arises from a cancellation of power laws, the IRFP seems to exhibit an intrinsically stronger topological character. The edge-selective structure of the critical distributions may suggest a boundary manifestation of the average Kramers-Wannier duality symmetry at the IRFP.

Electrowetting on textured and lubricant infused surfaces is conventionally expected to promote enhanced droplet spreading by reducing apparent contact angles. Contrary to this intuition, we report rapid tangential droplet ejection at applied DC voltages on specific microtextured, lubricant infused surfaces. Using high speed imaging and a precisely controlled electrowetting setup, we reveal the dependence of droplet dynamics on surface topology, wetting state, and the presence of a lubricant. On densely textured thick PDMS substrates of post spacing 5 to 10 um in a low hysteresis non-wetting Cassie state, and on all lubricant infused textured surfaces, droplets experience sudden lateral motion and eventual detachment. We attribute this counterintuitive phenomenon to unbalanced electrocapillary forces at the contact line combined with minimal pinning, which allows asymmetries in electric stresses to translate directly into net lateral motion. In contrast, Wenzel state droplets or surfaces with larger texture spacing exhibit conventional spreading with strong adhesion. By capturing the fundamental interplay among electrostatic driving forces, contact line pinning, and interfacial mobility, our results provide a new paradigm for controlled droplet transport and ejection in electrowetting systems mediated by dense micro posts and lubricant induced interfaces.

We study the mode dynamics of a generic quadratic fermionic Hamiltonian under a sudden quench protocol in momentum space. Modes with zero energy at any given time, $t$, are referred to as dynamical critical modes. Among all zero-energy modes, spin-flip symmetry is restored in the eigenvector corresponding to selected zero-energy modes. This symmetry restoration is used to define the dynamical quantum phase transition (DQPT). This shows that the occurrence of these dynamical critical modes is necessary but not sufficient for a DQPT. We show that the conditions on the quench protocol and time for such dynamical symmetry restoration are the same as the divergence of the rate function and integer jump in the dynamical topological order parameter, which have been the traditional identifiers of a DQPT. This perspective also naturally explains when one or both of DQPT and ground-state quantum phase transitions will occur.

We reformulate the neutral wealth tax framework of Froeseth (2026) in the language of stochastic dynamics and statistical physics. Individual wealth under geometric Brownian motion satisfies a Langevin equation with multiplicative noise; the probability density of wealth across a population then evolves according to a Fokker-Planck equation. A proportional wealth tax at market value enters as a uniform reduction of the drift coefficient, preserving the diffusion structure and all relative probability currents. This drift-shift symmetry is the physical content of tax neutrality. Each channel through which neutrality breaks down in practice, book-value assessment, liquidity frictions, forced dividend extraction, migration, and market impact, corresponds to a specific violation of this symmetry: a state-dependent, asset-dependent, or flow-dependent modification of the Fokker-Planck equation. The framework clarifies when wealth taxation is a benign rescaling of the dynamics and when it introduces genuinely new physics.

Focused ion beam irradiation of metastable Fe$_{78}$Ni$_{22}$ thin films grown on Cu(100) substrates results in the localized transformation of the originally paramagnetic, face-centered-cubic continuous film into ferromagnetic patterns with body-centered-cubic structure. The direction of the magnetic easy axis can be controlled by the focused ion beam scanning strategy, resulting in eight differently oriented crystallographic domains with different magnetic properties. We study the local crystallographic orientations of the transformed areas by electron backscatter diffraction and correlate these results with local magnetometry measurements. The observed magnetic anisotropy can be explained as a result of residual lattice strain after the fcc$\to$bcc transformation. These results extend the understanding of this material system and its transformation and allow for the patterning of high-quality magnetic nanostructures with precisely controlled magnetization landscapes.

Selecting the optimal resolution for discretizing high-dimensional data is a central problem in physics and data analysis, particularly in unsupervised settings where the underlying distribution is unknown. The Relevance-Resolution (Res-Rel) framework addresses this issue through an information-theoretic trade-off between descriptive detail and statistical reliability. Here we provide a systematic validation of this approach by comparing its characteristic optima--maximum relevance and the -1 slope (information-theoretic) point--with the discretization that minimizes the Kullback-Leibler divergence from a known or physically motivated ground truth distribution. Across unstructured and structured synthetic datasets, Gaussian clones of MNIST, and molecular dynamics simulations of the alanine dipeptide, we find that as the dimensionality or informative content increases the KL-optimal discretization consistently lies within the Res-Rel optimality region. Furthermore, in high-dimensional regimes the -1 slope criterion closely matches the KL divergence minimum. These results establish the quantitative consistency of unsupervised information-theoretic selection with distribution-based optimality.

The dramatic slowdown of dynamics in supercooled liquids approaching the glass transition remains one of the central unresolved problems in condensed matter physics. We review approaches that attribute this slowdown to growing thermodynamic or structural length scales and discuss their difficulties in accounting for recent numerical results. These limitations motivate the present review, which critically examines alternative theories in which the glassy slowdown is instead controlled by localized excitations and their elastic interactions. After reviewing key phenomenology with a focus on the fragility of liquids, dynamical heterogeneities, thermodynamics-dynamics correlation, and the effect of kinetic rules and swap algorithms, we compare elastic descriptions based on homogeneous and local heterogeneous elasticity to excitation-based theories incorporating nonlinear responses. Results are compiled to relate global and local elastic moduli, the Debye-Waller factor, and the density of excitations, leading to a quantitative theory testable in experiments. The thermal evolution of the excitation spectrum provides a parameter-free account of the activation energy, while their elastic interactions quantitatively reproduce dynamical heterogeneities via thermal avalanche processes. Synthesized together, these results lead to a framework where the evolution of the excitation spectrum, rather than the growth of a thermodynamic length scale, governs fragility in simple glass-forming liquids -- yet mean-field concepts of dynamical transitions remain central to describing excitations and building a real-space picture of relaxation.

Estimating entropy production in continuous systems that can only be observed with a limited resolution remains an open problem in stochastic thermodynamics. Extant estimators based on the measurement of waiting-time distributions require either the detection of Markovian events, which uniquely determine the state of the system, or assume a discrete underlying dynamics. We present a novel estimator that relies solely on the detection of a single particle leaving or entering regions, or crossing manifolds, in continuous space. This estimator is based on the frequency and the duration of transitions between such events. We derive this bound by introducing two kinds of discretization of space. Finally, we compare our novel bound to the TUR using simulations of a Brownian vortex and discuss its relation to other lower bounds to entropy production.

We train machine learning algorithms to infer the entanglement structure of disordered long-range interacting quantum spin chains by learning from the strong disorder renormalisation group (SDRG) method. The system consists of $S=1/2$-quantum spins coupled by antiferromagnetic power-law interactions with decay exponent $α$ at random positions on a one-dimensional chain. Using SDRG as a physics-informed teacher, we compare a Random Forest classifier as a classical baseline with a graph neural network (GNN) that operates directly on the interaction graph and learns a bond-ranking rule mirroring the SDRG decimation policy. The GNN achieves a disorder-averaged pairing accuracy close to one and reproduces the entanglement entropy $S(\ell)$ in excellent quantitative agreement with SDRG across all subsystem sizes and interaction exponents. RG flow heat maps confirm that the GNN learns the sequential decimation hierarchy rather than merely fitting final-state observables. Finite-temperature entanglement properties are incorporated via the SDRGX framework through a two-stage strategy, using the zero-temperature GNN to generate the RG flow and sampling thermal occupations from the canonical ensemble, yielding results in agreement with both numerical SDRGX and analytical predictions without retraining.

Altermagnetism has recently emerged as a new class of spin compensated magnetic materials that exhibit momentum dependent spin splitting despite having zero net magnetization. The origin of these electronic signatures lies in symmetry operations that connect opposite spin sublattices while allowing spin splitting in momentum space. While most candidate materials identified so far belong to inorganic crystals with fixed lattice symmetries, the realization of altermagnetism ultimately requires platforms in which magnetic symmetry can be deliberately engineered. In this Perspective, we discuss how metal-organic frameworks (MOFs) provide a unique chemical platform to address this challenge. We first place altermagnetism in the broader context of magnetic and electronically active metal-organic networks, highlighting how reticular chemistry enables precise control over lattice geometry, dimensionality and electronic structure. We then discuss how these features position framework materials as promising candidates for realizing altermagnetism and highlight the key challenges that must be addressed to translate theoretical proposals into experimentally accessible systems. Finally, we critically assess current experimental challenges and outline emerging directions for realizing and controlling altermagnetism in coordination framework materials, which emerge as a versatile and powerful platform for exploring new paradigms in spintronics.

Generative models based on invertible transformations provide a physics-aware route to sample equilibrium configurations directly from the Boltzmann distribution, enabling efficient exploration of complex thermodynamic landscapes. Here, we evaluate their applicability in regions where conventional simulations suffer from severe dynamical bottlenecks, focusing on the liquid-gas critical point of a Lennard-Jones fluid. We show that Boltzmann Generators capture essential signatures of critical behavior, retain reliable performance when trained at or near criticality, and extrapolate across neighboring states of the phase diagram. An intriguing observation is that the model's efficiency metric closely traces the underlying phase boundaries, hinting at a connection between generative performance and thermodynamics. However, the approach remains limited by the small system sizes currently accessible, which suppress the large fluctuations that characterize critical phenomena. Our results delineate the current capabilities and boundaries of Boltzmann Generators in challenging regions of phase space, while pointing toward future applications in problems dominated by slow dynamics, such as glass formation and nucleation.

Ferroelectric nematic phase (NF) represents an attractive and foremost field of liquid crystals, combining fluidity with ferroelectricity. NF materials exhibit large polarization values and remarkable non-linear optical properties. We have designed an original molecular structure with halogen substituents in the position of an electron donating group. In a prolonged molecular core, such a modification led to the presence of the ferroelectric nematic phase (NF) below the nematic one. Besides, an application of Cl atom in the molecular core of one of the presented materials has been utilized for the first time for ferroelectric nematogens. We have examined mesogenic behaviour and ferroelectric characteristics of the NF phase. In the NF phase for the cell with antiparallel rubbing, we have detected a textural transformation, which evidences strong polar character of anchoring at the surfaces. The presented results provide valuable insight into the design of ferroelectric nematic liquid crystalline compounds.

Edge transport serves as a powerful probe of remarkable low-energy properties of fractional quantum Hall states, including the anyonic character of their excitations. Here, we develop a theory of fractional quantum Hall edges driven out of equilibrium, which is based on the Keldysh action for the bosonized chiral Luttinger liquid. With this non-equilibrium FQH bosonization framework, we first consider a single-mode Laughlin edge and analyze the full counting statistics of charge, the quasiparticle Green's functions, and tunneling transport properties through a quantum point contact, allowing for generic edge excitations. We then extend the formalism to multi-mode edges with inter-mode interactions, and explore, with focus on the $ν=4/3$ and $ν=2/3$ edges as paradigmatic examples, how interaction-induced fractionalization of anyons modifies the edge dynamics and the associated transport observables. While the full counting statistics probes the fractionalized charge of the excitations, the Green's functions and tunneling transport are governed by mutual braiding phases of fractionalized excitations and tunneling quasiparticles. We emphasize in particular the effect of interaction-induced fractionalization on the Fano factor $F$ and the differential Fano factor $F_d$, observables that can be measured experimentally. Our formalism, which provides a unified framework for non-equilibrium transport in FQH edges and Luttinger liquids, permits extracting anyonic braiding information from non-equilibrium edge-transport experiments, and paves the way to various extensions, including more involved experimental geometries and edge structures.

Electrolyte solutions at high concentration are indispensable and yet poorly understood. In particular, the extent of speciation -- the formation of complexes composed of multiple species -- in concentrated ionic solutions is very challenging to obtain theoretically and experimentally, but can have a strong effect on solution properties. The literature is rife with contradictory estimates of speciation from experiments. We find that speciation affects transport properties, and is therefore, a prerequisite to accurately model concentrated solutions. We turn this to our advantage by showing that the viscosity can be used to determine the extent of complexation in concentrated aqueous solutions. Results of simulations as well as experimental measurements are presented. The atomistic Madrid-2019 force-field is extended to model FeCl$_2$. Solutions of FeCl$_2$ and MgCl$_2$ are compared and the observed difference in viscosity explained by more complexation in the former, a conclusion supported by recently reported X-ray absorption and neutron scattering experiments.

We develop a perturbative framework to calculate the mean-squared displacement (MSD) of active Brownian particles (ABPs) with a finite moment of inertia. Starting from the corresponding Fokker-Planck equation, we employ a Fourier transform for the spatial coordinates and Hermite polynomials as eigenfunctions for the angular velocity, which enables a systematic perturbative expansion of the MSD order by order. By resumming the resulting series in Laplace space and performing the inverse transform, we obtain an explicit expression for the MSD as a function of the moment of inertia. The analytical results are further validated by comparison with numerical simulations.

Symplectic quantization is a functional approach to quantum field theory that allows sampling of quantum fluctuations directly in Minkowski space time by means of a generalized Hamiltonian dynamics in an extra time variable $τ$ which, at large times, samples a microcanonical ensemble. In a previous work we showed that, for an interacting scalar theory in 1+1 dimensions, this framework captures genuine real time features that are inaccessible to Euclidean simulations. That original formulation suffers from two structural limitations, an ill defined non interacting limit and the lack of a direct correspondence between its correlation functions and those generated by the Feynman path integral. To solve these problems we introduced constrained symplectic quantization, a holomorphic reformulation in which fields and action are analytically continued and constraints are imposed on the intrinsic time Hamiltonian flow. The constraints select stable deterministic trajectories and they define convergent holomorphic integration cycles for the corresponding microcanonical measure. In the continuum limit we establish exact equivalence with the Feynman path integral at the level of the generating functional, thus providing a direct link between intrinsic time correlators and real time Green functions. In this contribution, we apply the method to the quantum harmonic oscillator on a real-time 1-dimensional lattice. Testing various observables, we find agreement between numerical and exact results for one- and two-point functions, and we reconstruct characteristic real-time features such as an oscillatory propagator, the discrete energy-gap spectrum, and the evolution of eigenstate probability densities. These tests provide numerical evidence that constrained symplectic quantization can sample real-time quantum observables and offers a practical route beyond Euclidean-time importance sampling.

We investigate the interference of Laughlin quasiparticles (QPs) in the fractional quantum Hall regime that are stochastically injected into a Fabry-Pérot interferometer. We find that the effective Aharonov-Bohm (AB) phase accumulated along the interferometer loop acquires an additional contribution of $\sin(2πλ)/2$ per QP present on it, where $πλ$ is the QP exchange phase. This contribution originates from time-domain braiding processes associated with injected QPs passing the interferometer quantum point contacts. In the limit of symmetric QP injection, the tunneling current noise exhibits AB oscillations as a function of the total injected current, providing access to the exchange phase $πλ$. In the regime of large total injection, we identify a universal Fano factor that displays power-law scaling and a characteristic phase shift reflecting real-space QP braiding along the interferometer edges. These results are relevant for accessing anyonic exchange statistics in mesoscopic interferometers.

The topological magnetoelectric effect (TME) in three-dimensional topological insulators (TIs), described by $ΔP = \frac{e^2}{2h} N_{\rm Ch}^{(2)} ΔB$, serves as a condensed-matter realization of the four-dimensional quantum Hall effect (4D QHE). In dual-gate axion-insulator devices, the TME-induced polarization yields a current $I_{\rm TME} \propto (C_{\rm total}/C_{\rm S})\,Q_{\rm 4D\text{-}QHE}$, where the signal is suppressed by the capacitance ratio $C_{\rm total}/C_{\rm S}$. Here we propose an active compensation scheme that introduces a tunable negative capacitance $C_{\rm comp} \approx -C_{\rm gate}$ into the gate line, effectively canceling the gate dielectric capacitance and driving $C_{\rm total}/C_{\rm S} \to 1$. We validate the method using a quantum anomalous Hall (QAH) device, which shares the same surface-state physics with the axion insulator but permits direct charge measurement via a single gate, recovering over $95\%$ of the quantized charge signal from an initially half-attenuated state. This compensation method provides a robust means of resolving minute TME signals, offering a promising pathway toward direct measurements of the 4D QHE.

Phases that go beyond dipolar ordering and into multipolar ordering have recently been observed in magneto-electric materials. The resulting phase diagram is commonly explained using the concept of competing orders and exact microscopic interactions. In contrast, we propose an approach based on composite order emerging from a parent phase to explain quadrupoling above magnetic or electric dipolar orders. We include thermal fluctuations and symmetry and show their influence on the emergence of quadrupolar order. We find an analytical expression for the quadrupolar transition temperature, the critical anisotropy and explain the coupling of the quadrupolar order to mechanical strain, in agreement with experiments. The shift in perspective on quadrupolar ordering from competing to composite order is universal and can be extended to other types of multipolar ordering. This offers the possibility of understanding tunability and material-specific predictions of the related phase transitions without explicit knowledge of the microscopic mechanisms.

We present a systematic investigation of superconductivity in a topological superconductor candidate $T_{\rm d}$-MoTe$_2$ in the few-layer limit. By examining multiple mechanically exfoliated samples with different thicknesses, substrates and crystal qualities, we quantitatively correlate superconducting temperature ($T_c$) with disorder, carrier density, carrier type and mobility. By integrating these experimental findings with first-principles calculations, we reveal the relationship between the band structure and superconductivity in this material. Notably, in 2 L samples we access a highly hole-doped regime that has not been systematically explored in previous experiments, providing a complementary perspective to earlier studies. In this regime, we demonstrate that superconductivity can be realized in a manner consistent with a conventional phonon-mediated $s_{(++)}$-wave pairing.

In order to describe the nonisothermal dynamics of two-phase flows or binary mixtures such as colloidal suspensions consisting of colloidal particles and solvent on a microscopic level, we derive a new extended dynamical density functional theory (EDDFT) that includes the total mass density, the local concentration of one species, the total momentum density, and the energy density as variables using the Mori-Zwanzig-Forster projection operator technique. Through the incorporation of the momentum density into EDDFT, not only the diffusive but also the convective dynamics is taken into account. We derive an exact entropy and free-energy functional for the case of hard spheres. The hydrodynamic limit of our new EDDFT and its relation to the mode-coupling theory of the glass transition are discussed. It is shown that EDDFT allows to obtain the correct value for the speed of sound.

Altermagnetism has emerged as a promising ingredient for realizing nontrivial Josephson phases, but so far explored in single Josephson junctions. In this work, we consider the coherent coupling of two Josephson junctions with spin-singlet $s$-wave superconductivity and demonstrate that $d$-wave altermagnetism gives rise to spin-polarized Andreev molecules due to the hybridization of Andreev bound states of each junction when the coupling is weak. Interestingly, these spin-polarized Andreev molecules induce an anomalous nonlocal Josephson effect, where the current flow across one Josephson junction due to phase changes across the other junction develops $0-π$ and $φ_{0}$ transitions originating from altermagnetism. Furthermore, the nonlocal Josephson current carried by spin-polarized Andreev molecules exhibits nonreciprocal critical currents, enabling a nonlocal Josephson diode effect whose polarity is tunable by the altermagnetic strength and right phase. Our findings put forward altermagnetism as a promising arena for designing nonlocal spin Josephson phenomena.

We report the absence of orbital Hall magnetoresistance (OMR) in nonmagnet/ferromagnet bilayers, challenging the general assumption that orbital transport mimics spin transport. Despite the observation of giant orbital torques, confirming the generation of orbital currents, thickness-dependent magnetoresistance measurements reveal that the signal is dominated by the intrinsic magnetoresistance of the ferromagnet and current shunting, with no discernible OMR contribution. We attribute this contradiction to the distinct transport properties of orbital compared with spin. Orbital currents undergo isotropic bulk absorption in the ferromagnet rather than anisotropic interfacial reflection required for OMR. Furthermore, we find that texture-induced magnetoresistance and self-torques in Ni-based bilayers can generate misleading signals, suggesting that caution is required when employing Ni in orbitronic studies. These findings clarify the distinct physical rules governing orbital transport and provide a simple method to distinguish spin and orbital currents.

We study the effects of spatially inhomogeneous diffusion on the non-equilibrium phase transition in the contact process. The directed-percolation critical point in the contact process is known to be stable against the addition of a spatially uniform diffusion term. Correspondingly, we find quenched randomness in the diffusion rates to be irrelevant by power counting in the field-theory of the contact process. However, large-scale Monte Carlo simulations demonstrate that such diffusion disorder destabilizes the clean directed percolation critical point. Instead, the transition belongs to the same infinite-randomness universality class as the contact process with disorder in the infection or healing rates. To explain these results, we develop an effective model with an infinite diffusion rate; it shows that diffusion disorder generates an effective disorder in the healing rates. The same mechanism also appears in the field-theoretic description: Whereas diffusion disorder is irrelevant by power-counting, it generates standard random-mass disorder under renormalization. We discuss the validity of this mechanism for other absorbing state transitions and non-equilibrium phase transitions in general.

The Seebeck coefficient in liquids often reaches the mV/K range, yet its microscopic origin remains unclear due to the complexity of electrolyte systems. Here we propose a minimal electrostatic theory focusing on solvation entropy. Using the extended Born equation with temperature ($T$)-dependent dielectric constant ($\varepsilon$), we quantitatively reproduce the experimentally observed magnitude. The theory clarifies that large valence, small cationic radius, small dielectric constant, and large $\frac{d\varepsilon}{dT}$ are key factors for enhanced liquid Seebeck response.

The statistical mechanics of particles that populate indistinguishable energy states is explored. In particular, the mathematical treatment of the microstates differs from conventional statistical mechanics where the energy levels or states are universally treated as distinguishable, and differentiated by unique quantum numbers, or addressed by distinct spatial locations. Results from combinatorial counting problems are adapted to derive exact distribution functions for both classical and quantum particles at high degeneracy levels. Classical particles exhibit a definitive glass transition, similar to supercooled liquids where where the configurational entropy vanishes below a finite temperature $T_K$.

The thermodynamics of randomly quenched disordered Ising metamagnet has been studied by Monte Carlo simulations. The disorder has been implemented either by inserting nonmagnetic impurity or by uniformly distributed quenched random magnetic field. The staggered magnetisation ($M_s$) (calculated from the sublattice magnetisation) and the corresponding staggered susceptibility ($χ$) are studied as functions of the temperature ($T$). The antiferromagnetic phase transition has been found while cooling the system from the high temperature paramagnetic phase. The transition temperature(or pseudocritical temperature ($T_c$)) has been found to decrease as the concentration ($p$) of nonmagnetic impurity increased. The nonmagnetic impurity dependent staggered magnetisation has been found to show the scaling behaviour $M_sp^b \sim (T-T_c)p^a$ (with $a \cong -0.95$, $b \cong 0.09$ and $T_c \cong 4.45$) obtained through the data collapse. The zero temperature staggered magnetisation ($M_s(0)$) has been found to decrease linearly. The critical temperature($T_c$) is showing a linear ($T_c=mp+c$) dependence with the concentration ($p$) of nonmagnetic impurity. The antiferromagnetic phase transition has been found to take place at lower temperature for the higher value of the width ($s$) of the uniformly distributed quenched random field. The critical temperature ($T_c$) has been found to show the nonlinear dependence ($T_c=a+bs+cs^2$) on the width ($s$) of the uniformly distributed random magnetic field. The extrapolation (both for $p \to 0$ and $s \to 0$) restores the Neel temperature of three dimensional pure Ising antiferromagnet.

The superconductor-insulator transition as a paradigm of quantum phase transitions has attracted tremendous interest over the past three decades. While the magnetic field and carrier density can be tuned to drive the transition, the role of disorder in the transition is not well understood due to the complicated interplay between superconductivity and electron localization. In this work, we controllably introduce disorder in a two-dimensional high-temperature superconductor by depositing iron clusters onto the superconducting monolayer Fe(Te,Se) crystalline film. The spectral evolution from superconducting gaps to insulating gaps with increasing disorder is detected by scanning tunneling spectroscopy measurements. When the disorder is strong, large U-shaped gaps are observed and attributed to the localization-enhanced Cooper pair correlation. Our observations provide the insight into the emergent phases of low-dimensional and high-temperature superconductors with disorder.

Quantum sensors based on individual spins provide unprecedented access to local magnetic fields in condensed matter, chemistry, and biology, with solid-state defect spins emerging as the leading platform. However, their molecular-sensing capabilities are limited by confinement to a host lattice, which prevents placement in close proximity to a target molecule. Molecular spins offer an alternative, enabling chemical tunability and flexible positioning relative to the target system. Here we present a nanoscale sensing platform that combines molecular electron spins, ultrasensitive mechanical readout, and Hamiltonian engineering. Using a modified XYXY dipolar decoupling sequence, we suppress electron-electron dipolar interactions across a broad distribution of control fields, extending coherence times to $\sim 400~μ$s in an attoliter-scale droplet containing $\sim$100 trityl-OX063 radicals. Leveraging this sequence, we demonstrate frequency-selective detection of nanotesla-scale AC fields and perform sensing and spectroscopy of small, local nuclear-spin ensembles. Collectively, these results establish SQUINT (Spin-based QUantum Integrated Nanomechanical Transduction) as a framework for quantum sensing that affords molecular-level control over sensor properties and enables direct integration into complex molecular targets.

Quadrupole ordering with the ordering wavevector at the L points in PrIr$_2$Zn$_{20}$ under magnetic fields is analyzed using classical Monte Carlo simulations based on an effective $Γ_3$ quadrupole model on the diamond structure. We demonstrate that competition between the magnetic field and quadrupole anisotropy leads to a rich phase diagram for magnetic fields applied parallel to [001], which includes switching between a single-q state and a double-q state. We also show that a symmetry-allowed biquadratic intersite interaction, corresponding to a hexadecapole interaction, is crucial for reproducing the weak-field topology observed in experiments.

Natural filaments, such as proteins, plant tendrils, octopus tentacles, and elephant trunks, can transform into arbitrary three-dimensional shapes that carry out vital functions. Their shape-morphing behavior arises from intricate patterning of active and passive regions, which are difficult to replicate in synthetic matter. Here, we introduce a filament-centric strategy for programmable shape morphing in which intrinsic curvature and twist are directly encoded within multimaterial elastomeric filaments during fabrication. By harnessing rotational multimaterial 3D printing (RM-3DP), we directly prescribe the filament's natural curvature--twist field $\mathbf{k}(s)$ through controlled material distribution and helical liquid crystal mesogen alignment. When heated above their nematic-to-isotropic transition temperature ($T_\mathrm{NI}$), the helically aligned LCE regions contract along their local director field, while passive regions remain essentially unchanged. This approach enables independent control of bending and torsion at every cross-section along the filament centerline: the principal natural curvatures of the filament along two orthogonal axes as well as the local twist. Next, we printed architected lattices composed of unit cells formed by sinusoidal filaments that either reversibly contract, expand, or exhibit out-of-plane deformations. Discrete elastic rod simulations of Janus filaments with different natural curvatures and twist, which are interconnected within the printed lattices, allow accurate prediction of their observed shape-morphing behavior. By integrating active-passive elastomers, additive manufacturing, and computational modeling, we have created shape-morphing matter with complex programmable responses for applications that rely on adaptive, robotic, or deployable architectures.

We investigate the formation and interaction of solitons in the non-integrable Ostrovsky equation characterized by anomalous (positive) dispersion. This equation is relevant for describing wave phenomena in various media, including plasma, solids, and optical fibers. Our findings indicate that certain Ostrovsky solitons, which possess zero total ''mass'' and exhibit non-monotonic asymptotic behavior, can arise from initial perturbations of a pulse-like nature. These solitons may organize into regular trains, where they are arranged according to their amplitude, or they may form irregular, nonstationary configurations of bound interacting solitons, or even multi-soliton structures. Furthermore, we demonstrate that the interactions between solitons in the Ostrovsky equation are inelastic, resulting in the emergence of a dominant soliton, or ''soliton-champion,'' within closed systems, such as those with periodic boundary conditions. In such systems, the soliton with the highest amplitude acts as a ''terminator,'' annihilating smaller amplitude solitons and absorbing a portion of their energy. We also analyze the recurrence phenomenon and determine that, although it resembles that of the Korteweg-de Vries equation, it exhibits distinct features.

The classical Darboux system governing rotation coefficients of three-dimensional metrics of diagonal curvature possesses an equivalent formulation as a sixth-order PDE for a scalar potential (related to the corresponding $τ$-function). We demonstrate that this PDE is Lagrangian and can be viewed as an explicit scalar form of the `generating PDE of the KP hierarchy' as discussed recently in Nijhoff [arXiv:2406.13423] in the Lagrangian multiform approach to the Darboux and KP hierarchies. Scalar Lagrangian formulations for differential-difference and fully discrete versions of the Darboux system are also constructed. In the first three cases (continuous and differential-difference with one and two discrete variables), the corresponding Lagrangians are expressible via elementary functions (logarithms), whereas the fully discrete case requires special functions (dilogarithms). Remarkably, dispersionless limits of the above Lagrangians provide a complete list of 3D second-order integrable Lagrangians of the form $\int f(u_{xy}, u_{xt}, u_{yt})\, dxdydt$.

Acoustic scattering by vehicle surfaces can have significant effects on overall noise levels. In this paper, we present a space-time Galerkin time-domain boundary element method (TDBEM) that offers several distinct advantages over contemporary scattering methods for prediction of acoustic scattering and shielding of complex aeroacoustic sources such as propellers and rotors. The time-domain approach allows efficient simulation of transient, rotating, and broadband noise sources, while the Galerkin formulation is robust and unconditionally stable without any tuned numerical parameters. The main challenge of the Galerkin approach, namely the numerically difficult double space-time integration, is resolved through an efficient decomposition-based quadrature procedure. We present three cases with analytical solutions to validate the method and study its numerical properties, demonstrating excellent agreement for scattering and shielding by a variety of different geometries. We then apply the TDBEM to a trailing edge-mounted propeller case, comparing the numerical predictions with experimental measurements. The results demonstrate good agreement between predicted and measured scattering and shielding in a practical application case.

We have successfully developed and implemented an entire and compact table-top high-order harmonics generation (HHG) setup from monochromatic and intense femtosecond ($10^{-15}$ s) laser pulses launched in a target composed of a high-purity monoatomic noble gas specie, which can be Argon or Helium, distinctively. Its frequency arrangement is distributed both in the full eXtreme UltraViolet (XUV, $22-124$ eV) spectral region and in the bottom part of the Soft-X Ray range (SXR, $124-132$ eV), at once. Specifically, the core of this coherent secondary light source is based solely on a homemade, modular, affordable, though sturdy, design. We take advantage of this opportunity to present our design guidance of the XUV generation from a hollow capillary waveguide apparatus, and our simple recipe regarding the alignment process of the latter, which is easily carried out thanks to our adjustable design. Then, a comprehensive description of our entire XUV beamline is described, and participate in adding essential contents to the existing literature. Concurrently, we conducted theoretical studies, in order to anticipate or explain our experimental results. Overall, we found very good consistency between the experimental and cost-effective time-consuming numerical results. Finally, our setup provides very good vacuum performance under high gas load pressures, to a few atmospheres. All of these attributes fulfill the requirements regarding ultrafast time-resolved pump-probe configuration in table-top element-sensitive spectroscopy of complex and integrated optoelectronic devices made of magnetic materials.

Spin-lattice relaxation theory has been developed over almost a century, but some cardinal assumptions on the nature of the interactions involved have never been fully verified. This includes the weak coupling approximation, which makes it possible to describe spin dynamics perturbatively and leads to the canonical description of spin relaxation in terms of one- and two-phonon processes. Here, we extend the first-principles theory of spin relaxation to three-phonon processes and apply it to the vdW crystal of a spin-1/2 Chromium nitride complex. Results show that three-phonon contributions to spin relaxation only become relevant at temperatures inaccessible to experiments for this molecule, thus providing unprecedented evidence for the validity of the weak spin-phonon coupling assumption in spin relaxation theory. At the same time, we numerically show that a relatively small increase in spin-phonon coupling would lead to a crossover between three- and two-phonon processes' efficiency at room temperature, illustrating the possibility for three-phonon effects in molecular materials as well as paving the way to a systematic exploration of strong coupling in spin systems.

The magneto-optic Kerr effect (MOKE) is often applied as a tool for the magnetic characterization of thin films. Here, the change in polarization upon reflection from the magnetized sample is mainly regarded as being linearly proportional to the magnetization $\mathbf{M}$ (LinMOKE). MOKE contributions of second order in $\mathbf{M}$, also known as quadratic MOKE (QMOKE), which are proportional to $\mathbf{M}^2$, have also been studied in the past and used in thin film characterization. Recently, we reported on a systematic investigation of third-order MOKE contributions, named cubic MOKE (CMOKE) in Ni(111) thin films. This CMOKE manifests itself as an anisotropic contribution to the MOKE signal (with regard to the crystallographic orientation) measured in longitudinal or transversal configuration in full magnetic saturation. While LinMOKE (odd in $\mathbf{M}$) and QMOKE (even in $\mathbf{M}$) can easily be separated by methods based on magnetization parity, this no longer holds true for LinMOKE and CMOKE (odd in $\mathbf{M}$). It is therefore crucial to be aware of CMOKE contributions in order to correctly interpret MOKE data. Here, we report on the observation of CMOKE in thin film heterostructures with structurally twinned and untwinned Co(111) layers, demonstrating that a large CMOKE is not only present in Ni thin films. Additionally, we show that the observed anisotropic contributions cannot stem from LinMOKE by analyzing their dependence on the angle of incidence (AoI) of light. While the QMOKE is almost vanishing in Co(111) using light with wavelengths of 635\,nm and 406\,nm, the CMOKE contributions reach up to about 30\% of the LinMOKE contribution at an AoI of 45 degrees and become even more dominant towards normal AoI, which emphasizes the importance of higher-order MOKE effects in magneto-optic experiments.

We assess two widely used scissors-correction implementations for first-principles calculations of second-harmonic generation in ultraviolet nonlinear-optical (UV-NLO) crystals, namely scheme-L [Phys.\ Rev.\ Lett.\ \textbf{63}, 1719 (1989)] and scheme-N [Phys.\ Rev.\ B \textbf{72}, 045223 (2005)]. To enable controlled and numerically robust comparisons, we derive a unified static-limit formulation that avoids spurious divergences and is applicable to both schemes, extending earlier static-limit treatments that were effectively restricted to scheme-L. Benchmark calculations for representative borate and phosphate UV-NLO crystals show that both schemes largely preserve the spectral line shape while mainly rescaling the overall response. Scheme-N yields systematically 15\%--25\% larger magnitudes than scheme-L, although for several tensor components scheme-L agrees better with available experimental benchmarks. We further show that Kleinman symmetry is satisfied in the static limit at the level of the formal theory, whereas apparent violations in practical calculations arise mainly from the sum-rule-based approximation used to evaluate generalized derivatives.

High harmonic generation (HHG) in solids results from an interplay between intraband acceleration and electron-hole recombination driven by a high-intensity laser pulse. Here, we theoretically reveal that the driving pulse duration can play a major role in extending HHG to higher photon energies by promoting higher conduction band excitations. The effect is present in a conventional semiconductor as Si, restricted in a large-gap insulator as MgO, and most prominent in RhSi, a prototypical chiral Weyl semimetal presenting numerous band crossings. Further, we elucidate the HHG selection rules in RhSi required for the synthesis of attosecond locally chiral light. The chiral crystal structure enables the generation of a local 3D electric field exhibiting an asymmetric instantaneous torsion on attosecond timescales. A pronounced circular dichroism emerges when the driving helicity is either aligned with or opposite to the crystal handedness. Our findings motivate future experiments in chiral Weyl semimetals to track high-energy band crossings and in-situ locally chiral light, paving the way for chiral compact light sources and light-wave driven topological electronics.

Froeseth (2026) shows that a proportional wealth tax on market values is neutral with respect to portfolio choice, Sharpe ratios, and equilibrium prices under CRRA preferences and geometric Brownian motion. This paper investigates the robustness of that result along two dimensions. First, we extend the neutrality frontier: portfolio neutrality, including all intertemporal hedging demands, is preserved under stochastic volatility (Heston and general Markov diffusions) and Epstein-Zin recursive utility, but breaks under non-homothetic preferences such as HARA. Second, we identify four channels through which implemented wealth taxes depart from neutrality even under CRRA: non-uniform assessment across asset classes, general equilibrium price effects in inelastic markets, progressive threshold structures, and endogenous labour supply. Each channel is formalised and, where possible, calibrated to the Norwegian wealth tax system. The progressive threshold introduces a tax shield that increases risk-taking near the exemption boundary, an effect opposite in sign to the HARA distortion, and, at the extreme, generates a participation margin at which investors exit the tax jurisdiction entirely. We formalise this tax-induced migration as the extreme response at the progressive threshold and examine the Norwegian post-2022 experience as a case study. The full framework is applied to evaluate the Saez-Zucman proposal for a global minimum wealth tax on billionaires and the related French proposal for a national minimum tax above EUR 100 million.

We analyse the effect of a proportional wealth tax on asset returns, portfolio choice, and asset pricing. The tax is levied annually on the market value of all holdings at a uniform rate. We show that such a tax is economically equivalent to the government acquiring a proportional stake in the investor's portfolio each period, a form of risk sharing in which expected wealth and risk are reduced by the same factor, while the return per share is unaffected. This multiplicative separability drives four main results: (i) the coefficient of variation of wealth is invariant to the tax rate; (ii) optimal portfolio weights are independent of the tax rate; (iii) the wealth tax is orthogonal to portfolio choice, inducing a homothetic contraction of the opportunity set that preserves the Sharpe ratio of every portfolio; (iv) taxed and untaxed investors price assets identically. Results are derived under geometric Brownian motion and generalised to the location-scale family. A Modigliani-Miller analysis confirms pricing neutrality and identifies an inconsistency in the literature regarding the discount rate for after-tax cash flows. Under CAPM with CRRA preferences, after-tax betas equal pre-tax betas and the security market line contracts by the tax factor; general-equilibrium prices are unchanged. This resolves an error in Fama (2021). The neutrality results depend on three conditions commonly violated in practice: universal taxation at market value, frictionless markets, and dividend consumption. We formalise three channels through which relaxing these conditions breaks neutrality: book-value taxation, liquidity frictions, and dividend extraction, and show they have opposing effects on asset prices.

Extreme values and the tail behavior of probability distributions are essential for quantifying and mitigating risk in complex systems of all kinds. In multivariate settings, accounting for correlations is crucial. Although extreme value analysis for infinite correlated systems remains an open challenge, we propose a practical framework for handling a large but finite number of correlated time series. We develop our approach for finance as a concrete example but emphasize its generality. We study the extremal behavior of high-frequency stock returns after rotating them into the eigenbasis of the correlation matrix. This separates and extracts various collective effects, including information on the correlated market as a whole and on correlated sectoral behavior from idiosyncratic features, while allowing us to use univariate tools of extreme value analysis. This holds even for high-frequency data where discretization effects normally complicate analysis. We employ a peaks-over-threshold approach and thereby fully avoid the analysis of block maxima. We estimate the tail shape of the rotated returns while explicitly accounting for nonstationarity, a key feature in finance and many other complex systems. Our framework facilitates tail risk estimation relative to larger trends and intraday seasonalities at both market and sectoral levels.

Urban mobility models are essential tools for understanding and forecasting how people and goods move within cities, which is vital for transportation planning. The spatial scale at which urban mobility is analysed is a crucial determinant of the insights gained from any model as it can affect models' performance. It is, therefore, important that urban mobility models should be assessed at appropriate spatial scales to reflect the underlying dynamics. In this study, we systematically evaluate the performance of three popular urban mobility models, namely gravity, radiation, and visitation models across spatial scales. The results show that while the visitation model consistently performs better than its gravity and radiation counterparts, their performance does not differ much when being assessed at some appropriate spatial scale common to all of them. Interestingly, at scales where all models perform badly, the visitation model suffers the most. Furthermore, results based on the conventional admin boundary may not perform so well as compared to distance-based clustering. The cross examination of urban mobility models across spatial scales also reveals the spatial organisation of the urban structure.

We report on the interaction between stimulated Brillouin scattering and temporal cavity solitons in doubly resonant ring resonators. Our experiments are performed in coherently driven passive optical-fibre resonators. We demonstrate that the interplay between four-wave mixing and cascade Brillouin lasing spontaneously generates patterns of CSs on a temporal grid at twice the Brillouin-shift. These patterns are shown to be highly stable owing to a long-range locking mechanism mediated by the acoustic oscillation generated by the solitons. We introduce a unified mean-field model of the cavity to describe the dynamics between the coupled forward and backward waves under coherent driving. This model reproduces very well the experiments and explains the paracrystalline structures of the soliton pattern. Our findings significantly advance the understanding of hybrid Brillouin-Kerr optical frequency combs.

We formulate and validate an equivalent circuit model describing mutual resistive coupling between three microwave cavity resonators interconnected via thin metallic foils. Each cavity is represented as a lumped LCR circuit, while the foils act as a dissipative interface that mediates energy exchange via mutual resistance. This coupling mechanism produces interference effects and a controllable anti-resonance when the input resonators are amplitude- and phase-balanced, a behavior not achievable with standard microwave antenna probes. All three resonators operated in the TM$_{010}$ mode, where two input resonators each excited the third via a thin copper foil. Analytical expressions are derived for the mutual resistance and coupling coefficient of these foils in this geometry. Under balanced conditions, a sharp anti-resonance emerges with a near order-of-magnitude enhanced phase sensitivity at the resonant frequency of the output cavity, consistent with model predictions. The experimentally extracted mutual coupling coefficients, $Δ_{13}=(5.00\pm0.01)\times10^{-6}$ and $Δ_{23}=(4.10\pm0.01)\times10^{-6}$, fall within the calculated range $Δ_{n3}\approx(1\text{--}48)\times10^{-6}$ derived from the foil's electromagnetic properties, where the spread is dominated by the estimated foil thickness uncertainty of $(9\pm1)\,μ\mathrm{m}$. These results confirm that resistive coupling can occur across a number of skin depths of a metallic interface, providing a new means of engineering controlled interference in multi-resonator systems. The approach offers potential applications in precision microwave experiments, phase-sensitive detection, and tests of fundamental electromagnetic interactions.

Microwave sensing is a critical enabler for all-weather perception, yet its resolution is fundamentally capped by the diffraction limit of the physical antenna aperture. While vortex electromagnetic (EM) waves offer a route to bypass this barrier, practical deployment is constrained by the trade-off between bandwidth, mode purity, and hardware complexity. Here, we propose a microwave photonic architecture enabled by a chip-scale dissipative Kerr soliton (DKS) microcomb that resolves these constraints. The microcomb provides a grid of over 270 optical lines with linewidths below 30 kHz, which are modulated and optically processed to synthesize vortex waves covering 8 GHz (18-26 GHz) with 15 programmable orbital angular momentum (OAM) modes. In contrast to conventional parallel-laser systems, our approach reduces phase error and improves OAM mode purity, while condensing the multi-wavelength source onto a monolithic chip. We demonstrate superior forward-looking imaging performance, clearly resolving both point targets and complex scenes. This work establishes a scalable framework bridging integrated soliton physics with broadband microwave processing, paving the way for next-generation compact smart sensors.

Magnonic crystals (MCs) are emerging spintronic metamaterials capable of manipulating transmission properties of magnons, the quanta of spin waves. Due to the complex relationship between lattice geometry and magnonic band dispersion, it remains challenging to establish general design strategies for optimizing targeted properties in MCs. In this study, we demonstrated an inverse-design framework for two-dimensional MCs to explore unconventional lattice structures with large magnonic band gaps. We employed genetic algorithms to enable global exploration of structures with a complete band gap as the objective property, and used frequency-domain micromagnetic simulations for computationally efficient band gap evaluation. Our established inverse-design method successfully discovered several previously unreported designs of MCs, whose performance was validated using time-domain micromagnetic simulations. Furthermore, we observed that the design landscape becomes increasingly non-convex at high-order bands, suggesting the existence of multiple design solutions. The overall inverse-design framework is expected to be broadly applicable to experimentally accessible material systems and device dimensions, facilitating the formulation of design rules for MCs.

Precise sensing and control of spatial mode content is essential for the performance of precision optical systems, particularly interferometric gravitational-wave detectors, where misalignment and mode mismatch can lead to significant optical losses and degraded quantum noise suppression. Conventional approaches, including heterodyne wavefront sensing and phase camera techniques, are effective but can be limited by hardware complexity and systematic uncertainties arising from restricted reference-beam overlap. This paper presents a novel two-step deep learning pipeline for robust beam diagnostics based solely on beam intensity images. In the first stage, a multi-intensity-image convolutional neural network (CNN) performs accurate mode decomposition, recovering the complex modal content of distorted beams. In the second stage, the predicted mode coefficients are fed into a downstream regression network that simultaneously estimates all eight degrees of freedom (DoFs) associated with misalignment and mode mismatch, including beam tilt, lateral offset, and waist size and position mismatches in both transverse directions. The proposed CNN-based framework achieves a mean absolute error (MAE) of 0.0034 in the mode decomposition stage, which propagates to a total MAE of 0.0062 in the recovered beam imperfection parameters at the final stage. This corresponds to an average residual optical loss of 39 ppm per DoF (310 ppm total). This approach relies only on standard CCD imaging and is robust to random intensity noise, eliminating the need for complex interferometric hardware. The results demonstrate that the proposed deep learning pipeline enables real-time, high-accuracy wavefront sensing and mode-mismatch diagnostics, providing a scalable and hardware-efficient tool for improving the stability and sensitivity of precision optical systems.

Atomically thin materials exhibit electronic and optical properties distinct from their three-dimensional counterparts. For metals, particularly gold, monolayer studies remain largely unexplored due to fabrication and characterisation challenges. Here we report the first optical study of a stable quasi-freestanding gold monolayer formed by Au intercalation between graphene and SiC. Mid-infrared nanoimaging reveals plasmon-polaritons with wavelengths nearly an order of magnitude shorter than free-space light. Analysis of their dispersion using a Drude model yields a relaxation time of $τ= 18\,$fs, comparable to bulk gold, and a Drude weight of $D = 1.3\,$mS$\cdot$eV, nearly twice the bulk expectation. These results establish monolayer gold as a two-dimensional metal, opening opportunities for nanoscale photonics, plasmonics and ultra-thin electronics.

The spatial public goods game has been used to examine factors that promote cooperation. Owing to the complexity of the dynamics of this game, previous studies on this model neglected analytical approaches and relied entirely on numerical calculations using the Monte Carlo (MC) simulations. In this paper, we present the approximate master equations (AMEs) for this model. We report that the results obtained by the AMEs are mostly qualitatively consistent with those obtained by the MC simulations. Furthermore, we show that it is possible to obtain phase boundaries analytically in certain parameter regions. In the region where the noise in strategy decisions is very large, the phase boundary can be obtained analytically by considering perturbations from the steady state of the voter model. In the noiseless region, discontinuous phase transitions occur because of the characteristics of the function that represents strategy updating. Our approach is useful for clarifying the details of the mechanisms that promote cooperation and can be easily applied to other group interaction models.

Perovskite based nanocomposites provide a versatile platform for tailoring light matter coupling and carrier injection in low dimensional optoelectronics. In particular hybrid structures incorporating transition metal dichalcogenides (TMDCs) enable tunable band alignment and excitonic dynamics. However, the central challenge lies in the scalable synthesis of defect controlled layered TMDCs suitable for coherent interfacial coupling. Here we demonstrate geometry dependent femtosecond (fs) laser ablation in liquid as a scalable strategy for defect controlled exfoliation of WS2 and in situ formation of WS2/CsPbBr3 nanocomposite. Using 50 fs pulses at identical fluence, we directly compare Gaussian and Bessel beam profiles to elucidate how spatial energy distribution governs nonlinear carrier generation, Coulomb stress and electron-lattice energy transfer. Multiphysics modeling combining nonlinear absorption, Coulomb instability criteria and two temperature dynamics reveals that the axially extended, weakly diffracting Bessel profile excites carrier densities above the electronic destabilization threshold while reducing the peak lattice heating. In contrast, Gaussian excitation concentrates energy within the focal volume, promoting rapid thermalization and defect formation. Spectroscopic analysis reveals reduced trap assisted recombination and enhanced excitonic stability in Bessel processed samples. Transient absorption spectroscopy (TAS) further demonstrates prolonged carrier lifetime. Extending this approach, single step ablation yields WS2/CsPbBr3 hybrids exhibiting Type-I band alignment and geometry dependent interfacial carrier injection, evidenced by photoluminescence quenching and modified ultrafast decay dynamics.

A multiband hybrid metasurface supporting multiple gap-surface plasmon (GSP) and localized surface plasmon (LSP) modes is presented. The structure adopts a metal-dielectric-metal configuration consisting of an aluminum bottom layer, a silicon dioxide spacer, and a bar-disc hybrid resonator patterned in the top aluminum layer. Optimized geometrical parameters yield four distinct resonances across the near-infrared and telecommunication bands, arising from the interplay between GSP modes and LSP excitations. The reflectance spectra are systematically analyzed as functions of geometric parameters and polarization, demonstrating tunable multiband operation. Experimental measurements of the fabricated metasurface show good agreement with numerical predictions. Furthermore, the second-harmonic generation (SHG) response is numerically investigated, revealing enhanced SH emission at the resonance wavelengths due to strong electromagnetic field confinement within the metal-dielectric-metal cavity. The proposed metasurface provides a compact platform for multiband and multifunctional nanophotonic applications.

Polarization is a problem in modern society. Understanding how opinions evolve through social interactions is crucial for addressing conditions that lead to polarization, consensus, or opinion diversity. Classical opinion dynamics models have explored bounded confidence and homophily, but most assume equal issue importance and purely attractive forces. We extend these frameworks by developing a stochastic agent-based model where individuals hold binary opinions on multiple issues of heterogeneous weights and interact through both attraction (with similar others) and repulsion (from dissimilar others). Our model reveals that the similarity threshold determining friend-or-foe interactions fundamentally shapes outcomes, which in this model can be of three types: consensus, polarization, and persistent pluralism, where each opinion combination occurs in the population. Low thresholds promote consensus, while high thresholds lead to polarization or persistent pluralism. Surprisingly, introducing even a single issue of arbitrarily small weight can destabilize stable states, thus changing the solution type and increasing convergence times by orders of magnitude. To explain these phenomena, we derive a deterministic system of ordinary differential equations and analyze equilibrium symmetries. For up to five-issue systems, we provide a complete characterization: all weight configurations fall into a number of cases, each exhibiting distinct symmetry cascades as the threshold varies. Our analysis shows polarization risk increases when importance concentrates on few issues. This suggests mitigation strategies: fostering cross-cutting social ties, broadening discourse beyond core issues, and introducing new topics to disrupt polarization. The symmetry-based framework reveals how issue salience and social tolerance jointly shape collective opinion evolution.

Compact models of memristors are essential for simulating large-scale neuromorphic systems, yet they often do not include description of complex dynamics like volatile relaxation and synaptic plasticity. We introduce a modular, computationally efficient memristor model that bridges this gap by integrating principles from physics and computational neuroscience. The model defines a framework consisting of a standard formulation of memristive device dynamics, a functional rule mapping state variables to cumulative conductance, a volatility module inspired by the theory of linear viscoelasticity and a saturation module implementing a linear-nonlinear technique. Additionally, we develop a formulation of synaptic-like plasticity inspired by a biological spike-timing-dependent plasticity (STDP) rule, which is compatible with the general framework for memristive devices. Finally, we propose a Laplace transform-based technique to derive the precise form of the mapping from state variables to cumulative conductance, replacing ad hoc voltage-current relationships with principled construction. We quantitatively validate the complete model against a rich set of experimental data from polymeric memristors exhibiting potentiation, synaptic-like plasticity and volatile decay. Our work presents a new paradigm for memristor modeling that is both practical for large-scale simulation and rich in explanatory power, providing a principled tool for the design of next-generation neuromorphic hardware.

Nonlinear Raman-Nath diffraction (NRND) and nonlinear Cherenkov radiation (NCR) are significant nonlinear diffraction phenomena in optics. Previous studies have primarily focused on NRND and NCR in uniaxial crystals, particularly in periodically poled lithium niobate (PPLN) crystals. However, research on these phenomena in biaxial crystals, such as periodically poled potassium titanyl phosphate (PPKTP), has been limited, and the study of NCR in PPKTP has not yet been undertaken.In this work, we experimentally investigated NRND and NCR phenomena in PPKTP crystals under varying incident angles, pump polarizations, poling periods, and crystal temperatures. Our findings indicate that PPKTP exhibits over ten times greater thermal stability compared to PPLN. This high thermal stability is promising for applications in parallel optical computing, as it helps reduce optical mode deviations and minimize bit error rates.

In-line digital holography (DIH) is a widely used lensless imaging technique, valued for its simplicity and capability to image samples at high throughput. However, capturing only intensity of the interference pattern during the recording process gives rise to some unwanted terms such as cross-term and twin-image. The cross-term can be suppressed by adjusting the intensity of reference wave, but the twin-image problem remains. The twin-image is a spectral artifact that superimposes a defocused conjugate wave onto the reconstructed object, severely degrading image quality. While deep learning has recently emerged as a powerful tool for phase retrieval, traditional Convolutional Neural Networks (CNNs) are limited by their local receptive fields, making them less effective at capturing the global diffraction patterns inherent in holography. In this study, we introduce HoloPASWIN, a physics-aware deep learning framework based on the Swin Transformer architecture. By leveraging hierarchical shifted-window attention, our model efficiently captures both local details and long-range dependencies essential for accurate holographic reconstruction. We propose a comprehensive loss function that integrates frequency-domain constraints with physical consistency via a differentiable angular spectrum propagator, ensuring high spectral fidelity. Validated on a large-scale synthetic dataset of 25,000 samples with diverse noise configurations (speckle, shot, read, and dark noise), HoloPASWIN demonstrates effective twin-image suppression and robust reconstruction quality.

Ultrafast laser writing of single lattice defects in wide-bandgap semiconductors is shown to present a new physical setting in which deeply subwavelength laser-writing positioning precision is attainable, but where the whole notion of positioning can only be understood in a statistical sense. We outline a framework for the analysis of this class of laser - matter interactions, grounding the concepts of optical super-resolution and subdiffraction positioning in statistical optics. Working along these lines, we derive closed-form solutions for physically meaningful quantifiers of laser-matter interactions on a subwavelength scale, suggesting a physically clear view of how deeply subdiffraction resolution can emerge from the interplay between determinism and stochasticity. We show that subdiffraction positioning precision in single-lattice-defect laser writing is achieved at the cost of a lower throughput, setting physical bounds on the scalability of integrated quantum photonic systems fabricated by means of super-resolving laser writing.

Hub removal in scale-free networks is known to degrade percolation robustness by raising the bond percolation threshold. We show that this same intervention simultaneously triggers a cascade phase transition under the Watts threshold model. In Barabási--Albert networks, removing the top 10\% of nodes by degree raises $p_c$ from 0.34 to 0.90, reducing robustness to random edge failure. Simultaneously, at cascade threshold $\varphi=0.22$, mean cascade size increases from $0.29\%$ to $20.6\%$, crossing from the subcritical to supercritical regime -- compounding, rather than trading off, the percolation degradation. Using a controlled experiment that independently varies hub presence and hub activation threshold, we demonstrate that hub-mediated cascade suppression is primarily dynamical: making hubs vulnerable without removing them produces 95\% cascades versus 19\% with removal. We operationalize Watts' (2002) stable-node mechanism as a quantitative condition $k > 1/\varphi$ and derive, under the configuration-model approximation, a closed-form expression for the post-removal cascade branching factor $z_1(\varphi, α, m)$. This predicts a newly opened interval where the pre-removal network is subcritical but the post-removal network is supercritical. We define $\varphi^*$ as the upper supercritical boundary, $\varphi^* = \sup\{\varphi : z_1(\varphi) \geq 1\}$. The $z_1$ derivation confirms that at $\varphi=0.22$, the pre-removal network is subcritical ($z_1=0.850$) while the post-removal network is supercritical ($z_1=1.195$). Our results establish that hub manipulation creates compounding vulnerability: both percolation and cascade metrics worsen simultaneously on static, unweighted networks. The effect vanishes in homogeneous networks (ER, WS).

Neighborhood algorithms may take a considerable percentage of computer time in discrete element methods (DEM). While the sort-and-sweep algorithm is ideal in some ways, as it only deal with particles whose relative positions change in one coordinate direction, the other directions must be processed too, for all particles. In contrast, tree-codes deal only with adjacent particles. We compare sort-and-sweep and tree-code neighborhood algorithms for two-dimensional DEM simulations of polygonal particles in a rotating drum with up to 12000 particles. We discuss the effects of system size and inlining on the performance with respect to the cache memory. For the tree code, the performance is slightly better, at the cost of significantly increased cyclomatic complexity. In particular, one benefit is improved possibilities for shared memory parallelization.

Structured light fields exploit spin and orbital angular momentum for precision manipulation, advanced imaging, and high-capacity communication. Orbital angular momentum coherent state beams interpolate between Hermite- and Laguerre-Gaussian beams, enabling continuous spatial control. We introduce a symmetry-based framework for joint control of polarization and spatial structure under the shared \$su(2)\$ Lie algebra of spin and orbital angular momentum. Within this structure, we construct total angular momentum fields as superpositions of circular polarization and Laguerre-Gaussian beams, and define their \$su(2)\$ coherent states within fixed-angular-momentum subspaces. A single complex parameter controls both polarization and spatial degrees of freedom, enabling continuous, symmetry-preserving tuning.

Orbital angular momentum (OAM), a topological degree of freedom of light, is theoretically invariant under continuous deformations; yet, its physical observability degrades precipitously in complex media, creating a fundamental "topology-observability gap." Here, we propose a novel paradigm for topological measurement based on the non-separable coupling between polarization and topological features in vectorial vortex beam. By constructing a topological non-separability measure derived from global Stokes fields, and integrating it with a physics-guided machine learning calibration framework that combines Bayesian Gaussian process regression with XGBoost-driven adaptive model selection, we achieve high-fidelity identification of topological features up to 200. Crucially, this robustness persists even when beam intensity and phase structures are completely distorted by extreme complex media, including strong atmospheric turbulence, oceanic turbulence, and high-temperature jet exhausts. This approach overcomes the dual bottlenecks of limited accessible OAM modes and susceptibility to perturbations that constrain conventional methods. Our work not only bridges the fundamental divide between topological theory and physical observability, but also establishes a robust framework for the reliable deployment of high-dimensional OAM in real-world complex environments, promising exciting advancements for wireless optical communications, remote topological sensing, and classical analogs of quantum information protocols.

Hippocampal neurons track positions of self, others, and gaze direction. However, it is unclear how their respective neural codes differ enough to avoid confusion while allowing for abstraction. We recorded from populations of hippocampal neurons while participants performed a joystick-controlled virtual prey pursuit task involving multiple moving agents. We found that neurons have mixed selective responses that map positions of self, prey, and predator, as well as gaze. Their codes occupied mostly orthogonal subspaces, but these subspaces geometric structure allowed them to be aligned by simple linear transformations. Moreover, their geometry supported generalization across spatial maps, such that a linear rule learned on one agent transfers to another. This scheme enables reliable individuation and abstraction across both agent identity and viewpoint. Together, these findings suggest that hippocampal spatial knowledge is structured as a family of geometrically related manifolds that can be flexibly aligned to different agents and gaze directions.

Maximum-likelihood (ML) detection in high-order MIMO systems is computationally prohibitive due to exponential complexity in the number of transmit layers and constellation size. In this white paper, we demonstrate that for practical MIMO dimensions (up to 8x8) and modulation orders, near-ML hard-decision performance can be achieved using a structured reduced-search strategy with complexity linear in constellation size. Extensive simulations over i.i.d. Rayleigh fading channels show that list sizes of 3|X| for 3x3, 4|X| for 4x4, and 8|X| for 8x8 systems closely match full ML performance, even under high channel condition numbers, |X| being the constellation size. In addition, we provide a trellis based interpretation of the method. We further discuss implications for soft LLR generation and FEC interaction.

Pinching antenna systems (PAS) have recently emerged as a promising architecture for flexible and reconfigurable wireless communications. However, their performance is fundamentally constrained by in-waveguide attenuation, which is non-negligible in practical dielectric waveguides and can severely degrade the achievable data rate, particularly for long waveguides. To overcome this limitation, we propose a dual-fed PAS (DF-PAS), in which each waveguide is equipped with two feed points located at the two ends, enabling dynamic feed-point selection based on user locations. This design effectively shortens the in-waveguide propagation distance and mitigates attenuation-induced power loss without modifying the waveguide structure or the PA actuation mechanism. We investigate the DF-PAS in both single- and multi-waveguide scenarios. For the single-waveguide case, we derive closed-form high-SNR approximations of the ergodic rate and obtain closed-form solutions for the optimal PA position and feed-point selection under time-division multiple access (TDMA). We then extend DF-PAS to a multi-waveguide scenario, where we first derive closed-form high-SNR approximations of the ergodic rate and then formulate a joint optimization problem over feed-point selection, PA placement, and beamforming under general orthogonal multiple access (OMA). To solve this problem efficiently, we develop a two-phase optimization framework that integrates greedy feed-point switching, gradient-based PA placement, and WMMSE-based beamforming. Simulation results demonstrate that the proposed DF-PAS consistently outperforms conventional single-fed PAS (SF-PAS) across various network configurations, validating its effectiveness as a practical and scalable solution for mitigating in-waveguide attenuation in PAS-enabled wireless networks.

Due to the conceptual simplicity, the linear filtering framework, notably the autoregressive (AR) process, has a long history in simulating clutter sequences with specified probability density functions (PDFs) and autocorrelation functions (ACFs). However, linear filtering inevitably distorts the input distribution, which may lead to inaccurate PDF reproduction or restrict applicability to very simple ACFs. To address these challenges, this study proposes a series-based analytic continuation strategy that revitalizes AR process clutter simulation by accurately precomputing the input pre-distortion required to compensate for AR filtering. First, the moments and cumulants of the AR input are derived based on the input-output relationship of the AR process, facilitating the moment and cumulant expansions of the Laplace transform (LT) and the logarithmic LT around zero, respectively. Second, both series expansions are analytically continued via the Padé approximation (PA) to recover the LT over the full complex plane. Notably, the PA-based continuation of the moment expansion, a conventional choice, can be highly inaccurate when the LT exhibits strong oscillations. By contrast, given the logarithmic LT generally has a simpler structure, the continuation of the cumulant expansion provides a more stable and accurate alternative. Third, the LT recovered from the cumulant expansion facilitates fast simulation of the AR input non-Gaussian white sequence via a random variable transformation method, thereby enabling an efficient AR process. Finally, simulations demonstrate that the proposed strategy enables accurate and fast simulation of non-Gaussian correlated clutter sequences.

Spectrally-efficient communication systems rely on the use of multi-level modulation formats. At the receiver side, a demodulator is often used to extract soft information about the transmitted bits. Such a demodulator is typically implemented in the digital domain. However, analog implementations of such demodulators are also possible. In this paper, we design and simulate an analog 8-ary pulse-amplitude modulation (8-PAM) demapper in IHP SG13G2 SiGe BiCMOS technology. We generalize and improve a design available in the literature for 4-PAM. A fully MOSFET-based 8-PAM design is proposed. Our simulations and design are completely based on open-source IC design tools. Our results show an energy efficiency of 0.33 pJ/bit for a data rate of 1Gbit/s.

Distributed multi-target tracking (DMTT) in limited field-of-view (FoV) sensor networks commonly suffers from label inconsistency, whereby different nodes disagree on the identity of the same target. Recent track-consensus DMTT (TC-DMTT) strategies mitigate this issue by enforcing kinematic and label agreement through metric-based track matching. Nevertheless, their behavior under adversarial conditions remains largely unexplored. In this paper, we reveal identity-level vulnerabilities in TC-DMTT and introduce the concept of label hijacking: an attack in which an adversary injects spoofed tracks to corrupt target identities across the network. Drawing on an analogy to classical pull-off deception in radar, we formalize a notion of attack stealthiness and derive an optimization-based strategy for crafting such attacks. A three-sensor network case study demonstrates the impact of the proposed attack on label consistency and tracking accuracy, showing successful target impersonation. Overall, this work highlights the need to rethink robustness at the consensus layer in DMTT frameworks.

This paper presents a method to derate the dependency on the decimation factor, $M$, of the pass-band droop inherent to $N$-th ordered comb decimators. It is achieved by cascading a symmetric 3-tap FIR filter in the integral stage of the corresponding comb decimator and choosing the coefficients only as a function of order $N$. The proposed derating method derived from the conventional comb decimator can be readily applied to any recently developed comb decimator and droop-compensation filter design method.

Radio Frequency Interference (RFI) is a growing concern for Global Navigation Satellite System (GNSS) reliability. The Cyclone GNSS (CYGNSS) constellation, designed for ocean wind retrieval via GNSS reflectometry (GNSS-R), provides Delay-Doppler Maps (DDMs) with noise floor metrics exploitable for spaceborne RFI detection. This study proposes a maximum-based DDM noise floor strategy that selects the highest noise floor value among four simultaneous GNSS reflections at each 0.5-second epoch, rather than their mean, preventing dilution of anomalous signals by unaffected channels. To suppress false alarms, a two-tier verification framework is introduced: (1) multi-satellite concurrent detection, confirming RFI when two or more CYGNSS satellites independently flag the same geographic region, and (2) temporal persistence verification, confirming a single-satellite detection only if threshold exceedance persists over a 10-second window. The physical basis for this criterion is established through slant-range geometry analysis between a ground-based jammer and the orbiting satellite. Performance is evaluated using CYGNSS Level 1 data from May 2025 in two regions: White Sands Missile Range, where NOTAM-announced GPS jamming tests were conducted, and the Middle East, where persistent RFI has been documented. The proposed method is compared against NASA's kurtosis-based RFI flags and a mean-based noise floor method. Results show that it detected RFI on three dates where the other methods produced negligible detections, and flagged 62% of total epochs in the Middle East compared to 46% (mean-based) and 33% (kurtosis-based). It also demonstrated capability to detect the early onset of gradually intensifying interference and atypical abnormal patterns not previously reported, highlighting the potential of maximum-based DDM noise floor analysis for sensitive and reliable spaceborne RFI detection.

Channel prediction has emerged as an effective solution for acquiring accurate channel state information (CSI) in the presense of channel aging. Existing methods have inherent limitations, with conventional Kalman filter (KF)-based approach being vulnerable to model mismatch and deep learning (DL)-based approaches producing overconfident predictions. To address these issues, we propose a DL-based conformal Bayes filter (DCBF) that integrates DL-based prediction, conformal quantile regression (CQR), and Bayesian filtering. The proposed framework enables principled fusion of calibrated priors and observations, yielding reliable channel predictions with the calibrated uncertainty. Simulation results demonstrate that DCBF significantly improves DL-based prediction and outperforms the KF-based method.

Nonlinearity in coherent fiber links is fundamentally driven by the temporal statistics and spectral structure of signal intensity. This paper develops a unified framework that links block-level energy statistics of shaped constellations to the low-frequency features of the intensity-fluctuation power spectral density (PSD), thereby enabling spectral-temporal co-design for nonlinear mitigation. A semi-analytical PSD model is derived for finitely block-shaped symbols (including Constant Composition Distribution Matching (CCDM) and Enumerative Sphere Shaping (ESS)), explicitly exposing contributions from self-beating dependent on symbol energy variance, inter-symbol beating dependent on mean symbol energy, and block-induced energy variance terms. A compact expression for the spectral-dip width is obtained that captures the block length, symbol rate, pulse roll-off, and chromatic dispersion. This yields design rules for lowering the low-frequency content. The low-frequency content most strongly drives the induced XPM. Resulting optimal symbol-rate laws are provided for shaped and unshaped systems, and are validated by Monte-Carlo simulations, which also confirm the distinct low-frequency behaviour of CCDM (suppressed DC) versus ESS (finite DC pedestal at moderate block lengths). The framework consolidates prior time- and frequency-domain views and supplies actionable guidance for choosing block length, symbol rate, and shaping method to reduce nonlinear interference in high-capacity WDM systems.

We find the dispersion relations of two elusive families of core-bound excitations of the Gross-Pitaevskii (GP) vortex, varicose (axisymmetric) and fluting (quadrupole) waves. For wavelengths of order the healing length, these two families -- and the well-known Kelvin wave -- possess an infinite sequence of core-bound, vortex-specific branches whose energies lie below the Bogoliubov dispersion relation. In the short-wavelength limit, these excitations can be interpreted as particles radially bound to the vortex, which acts as a waveguide. In the long-wavelength limit, the fluting waves unbind from the core, the varicose waves reduce to phonons propagating along the vortex, and the fundamental Kelvin wave is the only core-bound vortex-specific excitation. Finally, we propose a realistic spectroscopic protocol for creating and detecting the varicose wave, which we test by direct numerical simulations of the GP equation.

We introduce group surface codes, which are a natural generalization of the $\mathbb{Z}_2$ surface code, and equivalent to quantum double models of finite groups with specific boundary conditions. We show that group surface codes can be leveraged to perform non-Clifford gates in $\mathbb{Z}_2$ surface codes, thus enabling universal computation with well-established means of performing logical Clifford gates. Moreover, for suitably chosen groups, we demonstrate that arbitrary reversible classical gates can be implemented transversally in the group surface code. We present the logical operations in terms of a set of elementary logical operations, which include transversal logical gates, a means of transferring encoded information into and out of group surface codes, and preparation and readout. By composing these elementary operations, we implement a wide variety of logical gates and provide a unified perspective on recent constructions in the literature for sliding group surface codes and preparing magic states. We furthermore use tensor networks inspired by ZX-calculus to construct spacetime implementations of the elementary operations. This spacetime perspective also allows us to establish explicit correspondences with topological gauge theories. Our work extends recent efforts in performing universal quantum computation in topological orders without the braiding of anyons, and shows how certain group surface codes allow us to bypass the restrictions set by the Bravyi-K{ö}nig theorem, which limits the computational power of topological Pauli stabilizer models.

Continuous-variable quantum systems are central to quantum technologies, with Gaussian states playing a key role due to their broad applicability and simple description via first and second moments. Distinguishing Gaussian states requires computing their trace distance, but no analytical formula exists for general states, and numerical evaluation is difficult due to the exponential cost of representing infinite-dimensional operators. We introduce an efficient numerical method to compute the trace distance between a pure and a mixed Gaussian state, based on a generalized Lanczos algorithm that avoids explicit matrix representations and uses only moment information. The technique extends to non-Gaussian states expressible as linear combinations of Gaussian states. We also show how it can yield lower bounds on the trace distance between mixed Gaussian states, offering a practical tool for state certification and learning in continuous-variable quantum systems.

The realization of quantum error correction protocols whose logical error rates are suppressed far below physical error rates relies on an intricate combination: the error-correcting code's efficiency, the syndrome extraction circuit's fault tolerance and overhead, the decoder's quality, and the device's constraints, such as physical qubit count and connectivity. This work makes two contributions towards error-corrected quantum devices. First, we introduce mirror codes, a simple yet flexible construction of LDPC stabilizer codes parameterized by a group $G$ and two subsets of $G$ whose total size bounds the check weight. These codes contain all abelian two-block group algebra codes, such as bivariate bicycle (BB) codes. At the same time, they are manifestly not CSS in general, thus deviating substantially from most prior constructions. Fixing a check weight of 6, we find $[[ 60, 4, 10 ]], [[ 36, 6, 6 ]], [[ 48, 8, 6 ]]$, and $[[ 85, 8, 9 ]]$ codes, all of which are not CSS; we also find several weight-7 codes with $kd > n$. Next, we construct syndrome extraction circuits that trade overhead for provable fault tolerance. These circuits use 1-2, 3, and 6 ancillae per check, and respectively are partially fault-tolerant (FT), provably FT on weight-6 CSS codes, and provably FT on \emph{all} weight-6 stabilizer codes. Using our constructions, we perform end-to-end quantum memory experiments on several representative mirror codes under circuit-level noise. We achieve an error pseudothreshold on the order of $0.2\%$, approximately matching that of the $[[ 144, 12, 12 ]]$ BB code under the same model. These findings position mirror codes as a versatile candidate for fault-tolerant quantum memory, especially on smaller-scale devices in the near term.

Characterizing the dynamics of open quantum systems at the level of microscopic interactions and error mechanisms is essential for calibrating quantum hardware, designing robust simulation protocols, and developing tailored error-correction methods. Under Markovian noise/dissipation, a natural characterization approach is to identify the full Lindbladian generator that gives rise to both coherent (Hamiltonian) and dissipative dynamics. Prior protocols for learning Lindbladians from dynamical data assumed pre-specified interaction structure, which can be restrictive when the relevant noise channels or control imperfections are not known in advance. In this paper, we present the first sample-efficient protocol for learning sparse Lindbladians without assuming any a priori structure or locality. Our protocol is ancilla-free, uses only product-state preparations and Pauli-basis measurements, and achieves near-optimal time resolution, making it compatible with near-term experimental capabilities. The final sample complexity depends on linear-system conditioning, which we find empirically to be moderate for a broad class of physically motivated models. Together, this provides a systematic route to scalable characterization of open-system quantum dynamics, especially in settings where the error mechanisms of interest are unknown.

We study the decoding problem for quantum Tanner codes and propose to exploit the underlying local code structure by grouping check nodes into more powerful generalized check nodes for enhanced iterative belief propagation (BP) decoding by decoding the generalized checks using a maximum a posteriori (MAP) decoder as part of the check node processing of each decoding iteration. We mainly study the finite-length setting and show that the proposed enhanced generalized BP decoder for quantum Tanner codes significantly outperforms the standard quaternary BP decoder with memory effects, as well as the recently proposed Relay-BP decoder, even outperforming generalized bicycle (GB) codes with comparable parameters in some cases. For other classes of quantum low-density parity-check (qLDPC) codes, we propose a greedy algorithm to combine checks for generalized BP decoding. However, for GB codes, bivariate bicycle codes, hypergraph product codes, and lifted-product codes, there seems to be limited gain by combining simple checks into more powerful ones. To back up our findings, we also provide a theoretical cycle analysis for the considered qLDPC codes.

We present a fast and effective framework for analysing and designing syndrome-extraction circuits (SECs). Our approach is based on left-right circuits, a general design for SECs which maintain low depth by staggering $X$ and $Z$ checks without interleaving gates. Initially proposed for specific classes of codes, we generalise this construction to arbitrary CSS codes and optimise the circuit structure to achieve low qubit idling time, large effective distance, and reduced minimum-weight failure mechanisms. A key component of our framework is the formal notion of residual errors and their associated distance metrics, which form lightweight tools for capturing error propagation and quantifying the potential harm of circuit-level errors. Applying our automated framework to diverse classes of codes, we observe consistent improvements in logical performance of up to an order of magnitude compared to existing single-ancilla SEC designs. We also use these tools to prove that no non-interleaving SEC can achieve circuit distance $12$ for the gross code, and identify an explicit circuit that we conjecture achieves distance $11$, exceeding previously known constructions.

We investigate the quantum algorithm of Babbush et al. (arXiv:2303.13012v3) for simulating coupled harmonic oscillators, which promises exponential speedups over classical methods. Focusing on linearly connected oscillator chains, we bridge the gap between theory and implementation by developing and comparing three concrete realizations of the algorithm. First, we implement a sparse initial state preparation combined with product-formula (Suzuki-Trotter) Hamiltonian simulation. Second, we implement a fully quantum, oracle-based framework in which classical data are accessed via oracles, the Hamiltonian is block-encoded, and time evolution is performed using QSVT-based Hamiltonian simulation. Third, we propose an efficient alternative that combines the sparse state-preparation routine of the first approach with the oracle and block-encoding-based simulation pipeline of the second. We provide these implementations on Classiq, a high-level quantum design platform and provide appropriate resource benchmarks. Our simulation results show that the complex initial state preparation proposed by Babbush et al. can be circumvented at least in the linear-chain case. Finally, we illustrate two physical applications-extracting normal modes and simulating coarse-grained energy propagation-demonstrating how the algorithm connects to measurable observables. Our results clarify the resource requirements of the algorithm and provide concrete pathways toward practical quantum advantage.

Quantum-gas microscopes provide direct access to the phases of the Hubbard model, bringing microscopic insight into the complex competition between interactions, SU(2) magnetism, and doping. Alkaline-earth(-like) fermions extend this spin-1/2 paradigm by realizing higher symmetries and giving access to SU(N) Hubbard models, with rich phase diagrams to be unveiled. Despite its fundamental interest, a microscopic exploration of SU(N) quantum systems has remained elusive. Here we report the realization of a quantum-gas microscope for fermionic $^{87}$Sr. Our imaging scheme, based on cooling and fluorescence on the narrow intercombination line at 689 nm, enables spin-resolved single-atom detection. By implementing a spin-selective optical pumping protocol, we determine the occupation of each of the 10 spin states in a single experimental realization, a crucial capability for probing site-resolved magnetic correlations. We benchmark our method by observing single-particle Larmor precession across the full spin-9/2 ground-state manifold. These results establish $^{87}$Sr quantum-gas microscopy as a powerful approach to study exotic magnetism in the SU(N) Fermi-Hubbard model, and provide a new detection tool for studies in quantum simulation, computation, and metrology.

Amplitude estimation, in its original form, is formulated as phase estimation upon the Grover walk operator. Since its introduction, subsequent improvements to the algorithm have removed the use of phase estimation and introduced low-depth variants that trade speedup factors for lower circuit depth. We make the key observation that amplitude estimation is equivalent to estimating the energy gap of an effective Hamiltonian, whereby discrete time evolution is generated by amplitude amplification. This enables us to develop two amplitude estimation algorithms for both Heisenberg-limited and low-depth circuit regimes, inspired by statistical phase estimation techniques developed for seemingly unrelated early fault-tolerant ground-state energy estimation. Our approach has significant technical and practical benefits, and uses simplified classical post-processing compared to prior techniques -- our theoretical and numerical results indicate that we achieve state-of-the-art performance. Furthermore, while our approach achieves Heisenberg-limited scaling, we also establish optimal query-depth tradeoffs up to polylogarithmic factors in the low-depth regime with provable theoretical guarantees. Due to its flexibility, generality, and robustness, we expect our approach to be a key enabler for a broad range of early fault-tolerant applications.

Correlated noise is a critical failure mode in quantum error correction (QEC), as temporal memory and spatial structure concentrate faults into error bursts that undermine standard threshold assumptions. Yet, a fundamental gap persists between the stochastic Pauli models ubiquitous in QEC and the microscopic, non-Markovian descriptions of physical device dynamics. We close this gap by introducing \emph{Spatiotemporal Pauli Processes} (SPPs). By applying a multi-time Pauli twirl -- operationally realised by Pauli-frame randomisation -- to a general process tensor, we map arbitrary multi-time, non-Markovian dynamics to a multi-time Pauli process. This process is represented by a process-separable comb, or equivalently, a well-defined joint probability distribution over Pauli trajectories in spacetime. We show that SPPs inherit efficient tensor network representations whose bond dimensions are bounded by the environment's Liouville-space dimension. To interpret these structures, we develop transfer operator diagnostics linking spectra to correlation decay, and exact hidden Markov representations for suitable classes of SPPs. We demonstrate the framework via surface code memory and stability simulations of up to distance \(19\) for (i) a temporally correlated ``storm'' model that tunes correlation length at fixed marginal error rates, and (ii) a genuinely spatiotemporal 2D quantum cellular automaton bath that maps exactly to a nonlinear probabilistic cellular automaton under twirling. Tuning coherent bath interactions drives the system into a pseudo-critical regime, exhibiting critical slowing down and macroscopic error avalanches that cause a complete breakdown of surface code distance scaling. Together, these results justify SPPs as an operationally grounded, scalable toolkit for modelling, diagnosing, and benchmarking correlated noise in QEC.

An algorithm for the generation of shuttling sequences is necessary for the operation of a linear segmented ion-trap quantum computer. The present work provides an implementation of an algorithm that produces sequences proved to be optimal for circuits with a quantum Fourier transform-like structure. Such optimality was proved in previous work of our group. We first present an approach for qubit mapping, i.e. determining the initial ordering of the ions, termed the common ion order, and develop a heuristic algorithm for its implementation. We explain how this heuristic is integrated in the shuttling sequence generation algorithm described in the previous work. The results show the increased performance of the heuristic in terms of reducing the number of required shuttling operations. The number of ion displacements required exhibits a polynomial increase in terms of the number of qubits, such that these operations become the main contribution to the overall resource cost. Furthermore, we show that multiple zones for gate interactions can reduce the amount of qubit register reordering.

Quantum memory is a scarce and costly resource, yet little is known about which learning tasks remain feasible under severe memory constraints. We study the problem of computing global properties of quantum sequences when quantum systems must be measured individually, without storing or jointly processing them. In our setting, a bit string $x \in \{0,1\}^n$ is encoded into an $n$-qubit product state $|ψ_{x_1}\rangle \otimes \cdots \otimes |ψ_{x_n}\rangle$, and the goal is to infer $f(x) \in \{0,1\}$ from measurements of this quantum encoding. We consider a simple local strategy, which we call the greedy strategy, that applies the same optimal single-system measurement independently to each subsystem and then infers $f(x)$ from the outcomes. Our main result gives a complete characterization of when the greedy strategy is optimal: it achieves the same maximum success probability as an unrestricted global measurement if and only if the target Boolean function is affine (in all but finitely many cases). We establish a universal performance guarantee for general Boolean functions, showing that the success probability of the greedy strategy is always at least the square of the optimal global success probability, in direct analogy with the Barnum-Knill bound for the pretty good measurement. These results demonstrate that even under extreme memory constraints, simple local measurement strategies can remain provably competitive for learning global properties of quantum sequences.

Magic states are a scarce resource for two-dimensional qubit stabilizer codes. Magic state cultivation was recently proposed to reduce the cost of magic state preparation by measuring the transversal Clifford operator of the color code. Cultivation achieves $\sim 10^{-9}$ logical error rates for the $d=5$ color code, with substantially lower space-time overhead than magic state distillation. However, due to the $\mathcal{O}(d)$ depth of the Clifford measurement circuit, magic state cultivation becomes impractical for $d>5$. Here, we perform logical $XS^\dagger$ measurements on the color code by gauging a transversal Clifford gate, resulting in a constant-depth logical measurement circuit. We employ repeated gauging measurements with post-selection rather than performing error correction on the Clifford stabilizer code that emerges during the gauging protocol, thus gaining simplicity at the cost of scalability. Our protocol requires a regular square grid connectivity and yields logical error rates comparable to magic state cultivation. The $d=7$ version of our protocol gives access to the $10^{-12}$ logical error rate regime at $0.05\%$ physical error rate while retaining more than $1\%$ of the shots after the equivalent of the cultivation stage.

Cross-entropy benchmarking (XEB) with single-qubit reference sequences is widely used to characterize multi-qubit gates in large-scale quantum processors, despite the lack of a rigorous theoretical justification. Here we show that the commonly employed additive single-qubit errors approximation underlying this approach breaks down and leads to a systematic overestimation of gate fidelities. We derive an analytical expression for the joint decay of simultaneous single-qubit reference sequences and introduce a refined expression for the interleaved gate fidelity estimation. Experiments on a superconducting quantum processor validate the theory and demonstrate that fidelities obtained using XEB with single-qubit references agree with those extracted from standard interleaved randomized benchmarking (IRB), while achieving higher precision due to reduced reference-sequence errors. Our results establish theoretical foundation for the single-qubit-based XEB and show that, with appropriate post-processing, it enables a reliable and robust approach for entangling gates benchmarking without the need for multi-qubit Clifford reference sequences.

Quantum tracking control encodes the desired dynamics into a tailored driving field; here, we let the system find its own way there. We propose a real-time feedback control framework in which a proportional controller continuously corrects a simple transform-limited field based on the instantaneous mismatch between two systems' responses - producing the required control on the fly, without prior waveform design. The framework is demonstrated on two distinct examples: a single-active-electron atom, where hydrogen is driven to mimic argon's strong-field optical emission, and a Fermi-Hubbard chain, where a weakly interacting lattice reproduces the transport dynamics of a Mott-insulating reference. By shifting the control paradigm from predesigned inputs to adaptive response tracking, this approach establishes closed-loop feedback as a broadly applicable route to programmable quantum dynamics.

I reduce the cost to prepare magic states with lattice surgery operations on the surface code by using a recursive implementation of 15-to-1 magic state distillation. On a rotated surface code with distance $d$, $|T\rangle$ preparation requires a $d$-by-$3 d$ grid of data qubits for up to $15 d$ error correction cycles, and $|CCZ\rangle$ preparation requires a $3 d$-by-$2 d$ grid for up to $10.5 d$ cycles. However, a significantly lower physical error threshold than that of the underlying surface code is required to match the error probability of the output magic state with the logical error rate of the output surface code at large code distances.

Two-dimensional topological translationally-invariant (TTI) quantum codes, such as the toric code (TC) and bivariate bicycle (BB) codes, are promising candidates for fault-tolerant quantum computation. For such codes to be practically relevant, their decoders must successfully correct the most likely errors while remaining computationally efficient. For the TC, graph-matching decoders satisfy both requirements and, additionally, admit provable performance guarantees. Given the equivalence between TTI codes and (multiple copies of) the TC, one may then ask whether TTI codes also admit analogous graph-matching decoders. In this work, we develop a graph-matching approach to decoding general TTI codes. Intuitively, our approach coarse-grains the TTI code to obtain an effective description of the syndrome in terms of TC excitations, which can then be removed using graph-matching techniques. We prove that our decoders correct errors of weight up to a constant fraction of the code distance and achieve non-zero code-capacity thresholds. We further numerically study a variant optimized for practically relevant BB codes and observe performance comparable to that of the belief propagation with ordered statistics decoder. Our results indicate that graph-matching decoders are a viable approach to decoding BB codes and other TTI codes.

Quantum error correction is critical to the design and manufacture of scalable quantum computing systems. Recently, there has been growing interest in quantum low-density parity-check codes as a resource-efficient alternative to surface codes. Their adoption is hindered by the difficulty of compiling fault-tolerant logical operations. A key challenge is that logical qubits do not necessarily map to disjoint sets of physical qubits, which limits parallelism. We introduce clustered-cyclic codes, a quantum low-density parity-check code family with finite-size instances such as [[136,8,14]] and [[198,18,10]] that are competitive with state-of-the-art constructions. These codes admit a directly addressable logical basis, enabling highly parallel logical measurement layers. To leverage this structure, we propose parallel product surgery for quantum product codes. Using an auxiliary copy of the data patch and an engineered product-connection structure, the protocol performs many logical Pauli-product measurements in a single surgery round with small, fixed overhead. For clustered-cyclic codes, this yields surface-code-style maximal parallelism: up to k/2 disjoint Pauli-product measurements per round under explicit algebraic conditions. We prove that parallel product surgery preserves the code distance for hypergraph product codes and numerically verify distance preservation for the listed clustered-cyclic instances with k = 8. Finally, for the [[24,8,3]] clustered-cyclic code, treating half of the logical qubits as auxiliaries enables arbitrary parallel CNOTs on disjoint pairs; combined with symmetry-derived operations, these gates generate the full Clifford group fault-tolerantly.

The ability to fault-tolerantly prepare CAT states, also known as multi-qubit GHZ states, is an important primitive for quantum error correction. It is required for Shor-style syndrome extraction, and can also be used as a subroutine for doing fault-tolerant state preparation of CSS codewords. Existing approaches to fault-tolerant CAT state preparations have been found using computationally expensive heuristics involving SAT solving, reinforcement learning, or exhaustive analysis. In this paper, we constructively find optimal circuits for CAT states in a more scalable way. In particular, we derive formal lower bounds on the number of CNOT gates required for circuits implementing $n$-qubit CAT states that do not spread errors of weight at most $t$ for $1\leq t \leq 5$. We do this by using fault-equivalent rewrites of ZX-diagrams to reduce it to a problem of characterising certain 3-regular simple graphs. We then provide families of such optimal graphs for infinitely many values of $n$ and $t\leq5$. By encoding the construction of optimal graphs as a constraint satisfaction problem we find explicit constructions for circuits that match this lower bound on CNOT count for all $n\leq50$ and $t \leq 5$ and for nearly all pairs $(n,t)$ with $n\leq 100$ and $t\leq 5$ or $n\leq 50$ and $t\leq 7$, significantly extending the regimes that were achievable by previous methods and improving the resource counts for existing constructions. We additionally show how to trade CNOT count against depth, allowing us to construct constant-depth fault-tolerant implementations using $O(n)$ ancilla and $O(n)$ CNOT gates.

The usual belief propagation (BP) decoders are, in general, exchanging local information on the Tanner graph of the quantum error-correcting (QEC) code and, in particular, are known to not have a threshold for the surface code. We propose novel BP decoders that exchange messages on the decoding graph and obtain code capacity thresholds via standalone BP for the surface code under depolarizing noise. Our approach, similarly to the minimum weight perfect matching (MWPM) decoder, is applicable to any graphlike QEC code. The thresholds observed with our decoders are close to those obtained by MWPM. This result opens the path towards scalable hardware-accelerated implementations of MWPM-compatible decoders.

Simulating the dynamics of molecular excitons in complex nanophotonic environments requires integrating rigorous electromagnetic simulations with accurate treatments of open quantum system dynamics. In this work, we develop MQED-QD (Macroscopic Quantum Electrodynamics for Quantum Dynamics), a robust computational package for simulating exciton dynamics in arbitrary dielectric and plasmonic environments. Based on the MQED framework, the package offers a unified workflow for constructing the dyadic Green's functions from classical electromagnetic solvers, parametrizing quantum master equations, and propagating the time evolution to determine the molecular subsystem's dynamical properties. To demonstrate the package's capabilities, we simulate exciton transport within a one-dimensional molecular chain near a silver nanostructure, including benchmarking against planar surfaces and exploring the influence of silver nanorods. Our results reveal that surface plasmon polaritons on nanorods dramatically enhance long-range dipole-dipole interactions, accelerating exciton delocalization and yielding higher participation ratios compared to planar geometries. By elucidating accurate molecular exciton dynamics in conjunction with nanophotonics and plasmonics, MQED-QD provides a powerful, open-source package that facilitates the rational design of nanoscale architectures.

Nonreciprocal phenomena are currently a major focus of research within the fields of classical and quantum technology. In this work, we theoretically investigate the interplay among multiple magnomechanically induced transparency (MMIT) windows, Fano resonances, slow/fast light, and nonreciprocal absorption and group delay in a hybrid cavity magnomechanical system. This system is composed of two yttrium iron garnet (YIG) spheres and a membrane positioned at the center of the cavity. By analyzing the absorption spectrum of a weak probe field in the presence of a strong control field, we demonstrate the emergence of five transparency windows resulting from combined photon-phonon, photon-magnon, and phonon-magnon interactions. The photon-phonon coupling associated with the membrane plays a crucial role in enhancing and tailoring these transparency features. We further examine the impact of the Barnett effect on the absorption and dispersion characteristics, showing that it enables the controllable manipulation of transparency windows and the generation of tunable Fano resonance profiles. The influence of cavity decay and magnon dissipation rates on the spectral response is also analyzed. In addition, we demonstrate that the group delay of the transmitted probe field can be effectively tuned via the photon-phonon coupling strength and the Barnett effect, allowing for a controllable transition between slow and fast light regimes. Finally, nonreciprocal absorption and group delay are achieved through appropriate adjustment of the coupling parameters. These findings highlight the potential of the proposed hybrid system for applications in optical signal processing and quantum information technologies.

We present a quantum-classical hybrid implementation of the Liouvillian recursion method to compute many-body Green's functions using a quantum computer. From an approximate ground state preparation circuit, this algorithm produces the local ($r=r'$) and inter-site ($r\neq r'$) Green's functions $G_{rr'}(ω)$ by measuring observables generated recursively. We demonstrate the approach on a superconducting quantum processor for the open-boundary four-site Hubbard model. We then use the computed Green's functions as input to the Galitskii-Migdal formula to produce better ground state energy estimation than the expectation value of the Hamiltonian for the approximate circuit. Empirical results indicate exponential convergence in the number of iterations, yielding a computational complexity polynomial in the Green's-function accuracy, as measured with the Wasserstein distance. Our results also indicate significant robustness to noise and to inaccuracies of the ground state preparation, providing evidence that Liouvillian recursion is well adapted to the constraints of near-term quantum computing.

Quantum error correction will likely be essential for building a large-scale quantum computer, but it comes with significant requirements at the level of classical control software. In particular, a quantum error-correcting code must be supplemented with a fast and accurate classical decoding algorithm. Standard techniques for measuring the parity-check operators of a quantum error-correcting code involve repeated measurements, which both increases the amount of data that needs to be processed by the decoder, and changes the nature of the decoding problem. Knill error correction is a technique that replaces repeated syndrome measurements with a single round of measurements, but requires an auxiliary logical Bell state. Here, we provide a theoretical and numerical investigation into Knill error correction from the perspective of decoding. We give a self-contained description of the protocol, prove its fault tolerance under locally decaying (circuit-level) noise, and numerically benchmark its performance for quantum low-density parity-check codes. We show analytically and numerically that the time-constrained decoding problem for Knill error correction can be solved using the same decoder used for the simpler code-capacity noise model, illustrating that Knill error correction may alleviate the stringent requirements on classical control required for building a large-scale quantum computer.

Using a simple spacetime hosting a strong curvature naked singularity, we employ an analogue gravity model to study electromagnetic fields in this background. We find exact solutions to the full set of electrostatic and electrodynamic equations that remain regular even in the presence of the singularity. Moreover, certain solutions sustain a regular and bounded power flux across the singularity, suggesting that electromagnetic energy may be transmitted through it.

Recent advancements in quantum technologies have highlighted the importance of mitigating system imperfections, including parameter uncertainties and decoherence effects, to improve the performance of experimental platforms. However, most of the previous efforts in quantum control are devoted to the realization of arbitrary unitary operations in a closed quantum system. Here, we improve the algorithm that suppresses system imperfections and noises, providing notably enhanced scalability for robust and optimal control of open quantum systems. Through experimental validation in a superconducting quantum circuit, we demonstrate that our approach outperforms its conventional counterpart for closed quantum systems with an ultra-low infidelity of about $0.60\%$, while the complexity of this algorithm exhibits the same scaling, with only a modest increase in the prefactor. This work represents a notable advancement in quantum optimal control techniques, paving the way for realizing quantum-enhanced technologies in practical applications.

Using canonical quantisation, and eschewing the Schwinger-Keldysh path integral, we derive a version of the Worldline Quantum Field Theory (WQFT) formalism suitable for both scattering and bound configurations of the classical two-body problem. Focusing on a pair of charged particles interacting via a scalar field, we quantise Hamilton's equations both in flat space and around a non-zero background, perturbing in post-Lorentzian (PL) and self-force (SF) expansions respectively. Our quantisation procedure provides access to the Magnus series, and is perfectly suited for computing matrix elements of $\hat{N}(t,t_0)=- i \hbar \log\hat{U}(t,t_0)$, both with and without external scalar states, for finite time intervals (bound orbits) and infinite time intervals (scattering). Doing so, we provide a complete set of gauge-invariant matrix elements describing the 1SF scattering dynamics up to 3PL order, and corresponding matrix elements for bound orbits. We also demonstrate how $\hat{N}$-matrix elements encode physical observables, providing a unified operator-based framework for conservative and radiative dynamics of binary systems. The new WQFT formalism generalises naturally to both gravity and electromagnetism.

In light of recent interpretations attributing pulsar timing array (PTA) signal to second-order gravitational waves induced by linear cosmological curvature perturbations in the early universe, the overproduction of primordial black holes (PBHs) poses a theoretical tension. In this work, we address this issue through extending such a scalar-induced gravitational wave (SIGW) framework to include third-order gravitational waves, which allow for a substantial enhancement in the spectral amplitude of SIGWs. Analyzing a combined dataset from cosmic microwave background and baryon acoustic oscillations, we derive cosmological constraints on the physical energy-density fraction of cosmological gravitational waves. Further incorporating PTA data, we obtain constraints on the spectral amplitude and peak frequency of SIGWs. Our results indicate that the parameter region favored by the data combination can to some extent alleviate the PBH overproduction problem, thereby supporting the theoretical consistency of our model. Furthermore, we demonstrate the robustness of our SIGW interpretation for the PTA signal by extending the analysis to include a gravitational wave background from supermassive black hole binaries. These findings are poised for further scrutiny with future high-precision observations.

Optimization is ubiquitous in quantum information science and technology, however, the corresponding optimization landscape can encounter false traps, i.e., local but not global optima, likely to prevent used optimizers from finding optimal solutions. Such traps are believed to arise from parameter insufficiency and are expected to disappear when tunable parameters are sufficiently abundant. In this work, we investigate optimization landscapes of quantum optimization problems, and especially obtain that the parameter sufficiency is not enough to ensure the absence of false traps. First, we present a complete framework for analyzing critical features of optimization landscapes, by deriving necessary and sufficient conditions to identify all critical points and to classify them as local maxima, minima, or saddles, under some assumptions. Then, we show that false traps can still emerge on landscapes even with sufficient parameters, implying their appearance cannot be solely attributed to parameter insufficiency. Moreover, a close connection between landscape topology and quantum distinguishability is revealed that the emergence of false traps is linked to the loss of distinguishability among states or operators in the objective function. Finally, implications of our results are noted. Our work not only provides a deeper understanding of the intrinsic complexity of quantum-classical optimization, but also provides practical guidance for solving quantum-classical optimization problems, thus significantly aiding the progress in witnessing quantum advantages of the underlying quantum information processing tasks.

Kerr black hole (BH) superradiance can form gravitational atoms and produce characteristic gravitational-wave signals, providing a probe of ultralight bosons and dark matter. In realistic systems, accretion-disk gravity can shift energy levels and mix states, modifying the effective superradiant growth. We model the disk as a weak external perturbation via a multipole expansion and derive an effective three-level Hamiltonian for the $n=2$ subspace $\{\ket{211},\ket{210},\ket{21-1}\}$ in the weak-coupling regime. The leading disk effect is the quadrupolar ($\ell_d=2$) tidal field, whose symmetries fix the selection rules: axisymmetry gives only diagonal shifts, equatorial nonaxisymmetry activates $Δm=\pm2$ mixing ($\ket{211}\leftrightarrow\ket{21-1}$), and breaking equatorial reflection opens $Δm=\pm1$ couplings involving $\ket{210}$. As illustrations, a transient equatorial $m=2$ spiral wave drives the resulting two-level system and can suppress or quench superradiance by populating a decaying mode, while a quasi-static warp produces full three-level mixing and can generate narrow ``growth gaps'' near accidental near-degeneracies, with the same static reshuffling also allowing enhancement when weight shifts toward the growing mode. These findings demonstrate that accretion disk perturbations are a crucial environmental factor in determining the dynamics of BH superradiance and the evolution of boson clouds, thereby providing a more reliable theoretical basis for assessing the detectability of ultralight bosons in realistic astrophysical settings.

Unidimensional (1D) Gaussian-modulated continuous-variable quantum key distribution protocols have been proposed as a way to simplify implementation and reduce costs through single-quadrature modulation, requiring only one modulator while maintaining compatibility with standard optical infrastructure. Here, we determine security bounds for 1D discrete-modulated protocol under the Gaussian extremality assumption by extending the method of Ghorai et al. [Phys. Rev. X 9, 021059 (2019)]. We establish the appropriate symmetry arguments to extend the method to the 1D discrete-modulated case, define the physicality zone in which the protocol is allowed to operate, and prove security against collective attacks in the asymptotic regime via semidefinite programming. Our analysis for uniformly distributed coherent states reveals a fundamental limitation: the Gaussian extremality assumption systematically overestimates Eve's information with increasing constellation size, yielding bounds so conservative that secure key extraction becomes impossible for constellations larger than four states, even under ideal conditions. This overestimation worsens with excess noise and restricts viable modulation amplitudes to impractically small values. Unlike two-dimensional (2D) protocols, where Gaussian extremality improves with constellation size, 1D protocols lack the growing phase-space isotropy required for the approximation to remain tight as the constellation grows. Our results expose these limitations and highlight the necessity of alternative methods or optimized non-uniform constellation designs for this class of protocols.

Dynamic quantum circuits combine mid-circuit measurement with classical feed-forward, enabling circuit constructions with reduced entangling-gate depth. Here, we investigate their use in Quantum Imaginary Time Evolution (QITE), where circuit depth and parameter growth limit practical implementations of ground-state preparation. For dense classical optimization Hamiltonians, we introduce a reduced-parameter QITE ansatz that restricts entanglement generation via a small set of control qubits, enabling each QITE layer to be implemented with constant two-qubit gate depth using fan-out-based dynamic circuits. In noiseless simulations of exact cover and set partitioning instances, the reduced ansatz yields a higher success probability than standard QITE approaches. We implement unitary, dynamic fan-out, and semi-classical adaptive variants on IBM superconducting hardware. The semi-classical variant performs favorably to the unitary implementation, while the fully dynamic construction exposes the trade-offs between entangling-depth reduction and measurement and feed-forward overhead associated to dynamic circuit implementations. Using a fidelity threshold of 0.5 relative to the noiseless QITE ansatz, we show that dynamic fan-out based QITE would outperform unitary implementations on current devices when the measurement and two-qubit gate errors are reduced by 65% and the feedback latency is halved.

Classical tensor network and hybrid quantum-classical algorithms are promising candidates for the investigation of real-time properties of lattice gauge theories. We develop here a novel framework which enforces gauge symmetry via a quantum-link virtual rishon representation applied at intermediate steps. Crucially, the gauge and matter degrees of freedom are dynamical variables encoded in terms of qubits, enabling analysis of gauge theories in $d+1$ spacetime dimensions. We benchmark this framework in a U(1) gauge theory with and without matter fields. For $d = 1$, the multi-flavor Schwinger model with $1\leq N_f\leq3$ flavors is analyzed for arbitrary boundary conditions and nonzero topological angle, capturing signatures of the underlying Wess-Zumino-Witten conformal field theory. For $d = 2$, we extract the confining string tension in close agreement with continuum expectations. These results establish the virtual rishon framework as a scalable and robust approach for the simulation of lattice gauge theories using both classical tensor networks as well as near-term quantum hardware.

Understanding the fundamental limits of state convertibility is crucial for establishing the boundaries of quantum information processing and thermodynamic efficiency. While auxiliary systems, catalysts, can facilitate otherwise impossible transformations, standard catalysis rigidly requires the auxiliary system to return to its exact initial state. In this work, we investigate the power of flexible catalysis, where the catalyst evolves through a cycle of states, restoring its initial configuration only after a finite number of steps. Focusing on the regime of fixed, finite dimensions, we analyze the capabilities of flexible catalysis within the resource theories of entanglement and quantum thermodynamics. In the context of entanglement, we derive conditions limiting flexible catalysts and demonstrate that they offer a strict advantage in the success probability of stochastic local operations and classical communication. Conversely, in quantum thermodynamics, we prove that flexible catalysis strictly outperforms standard catalysis even in deterministic settings. We provide an example identifying state transformations that are impossible with any standard catalyst of fixed dimension and Hamiltonian but become achievable via a flexible cycle.

Recent progress in fault-tolerant quantum computing suggests that leveraging error-syndrome information at the logical layer can substantially improve performance, including the estimation of logical observables from noisy states. In this work, based on quantum estimation theory, we develop an information-theoretic framework to quantify the utility of error syndromes for noisy logical observable estimation. We distinguish two operational regimes of such syndrome-aware protocols: classical protocols, in which the logical measurement basis is fixed and syndrome information is used only in classical post-processing, and quantum protocols, in which the logical quantum control can be tailored to depend on the observed error syndrome. For classical syndrome-aware protocols, we prove a universal limitation: on average, syndrome information can improve the effective logical error rate by at most a factor of two, implying at most a quadratic reduction in sampling overhead. In contrast, once syndrome-conditioned quantum control is permitted, we exhibit settings in which the effective logical error rate decays exponentially with the number of logical qubits. These findings provide fundamental guidance for designing future fault-tolerant architectures that actively exploit syndrome records rather than discarding them after decoding.

With an increasingly diverse portfolio of quantum backends, the adoption of standardized interfaces has become a key prerequisite for scalable access and interoperability within quantum software stacks. The Quantum Device Management Interface (QDMI) addresses this challenge and is emerging as one of the de facto standards for hardware abstraction, enabling the unified management not only of individual Quantum Processing Units (QPUs) but also of complete full-stack cloud services. This paper presents a case study demonstrating the integration of QDMI with Amazon Braket, a quantum computing cloud service that provides a single access point to a wide range of hardware technologies. By treating the cloud service itself as a unified device, the proposed implementation enables management of the complete task lifecycle - ranging from authentication and circuit submission to result retrieval - across Braket's heterogeneous set of simulators and hardware backends. We detail the engineering insights gained from this integration and present a hands-on example workflow, ultimately paving the way for integrated access to cloud-hosted quantum resources from QDMI-enabled software stacks.

Classical shadow tomography has emerged as a powerful framework for predicting properties of quantum many-body systems with favorable sample complexity. Standard theoretical guarantees, however, rely on the assumption that experimental rounds are independent and identically distributed (i.i.d.). This idealization is often violated in practice, where parameter drift, environmental noise, and active feedback generate history-dependent sequences of states or channels. To address this, we introduce a robust classical shadow protocol based on a truncated mean estimator. We prove that its sample complexity for predicting properties of the time-averaged state or channel matches the standard i.i.d. scaling governed by the shadow norm, even when experimental rounds depend arbitrarily on the past. Our results establish the robustness of the shadow formalism beyond the i.i.d. regime.

We construct a renormalization-group improved Schwarzschild-like black hole geometry using the exact new scheme running for the Newton coupling. The scale identification is implemented via a standard interpolating proper-distance function that smoothly connects the ultraviolet and infrared regimes. We present the resulting coordinate-dependent coupling and the improved metric function, analyzing its asymptotic expansions. The large-distance limit is shown to recover the classical Schwarzschild solution, while the short-distance behavior exhibits a regular de Sitter-like core, demonstrating the regularization of the central singularity. We also analyze the thermodynamic properties of the solution, showing that quantum corrections significantly modify the small-radius behavior, leading to a remnant configuration and a nontrivial phase structure. Finally, we perform a topological classification of the thermodynamic phase space and demonstrate that asymptotically safe effects shift the critical point while preserving the global topological number of the Schwarzschild solution.

We numerically investigate the nonlinear dynamics of a two-dimensional exciton-polariton quantum fluid coherently driven by two counter-propagating laser beams. Using an exciton-photon coupled driven-dissipative Gross-Pitaevskii framework, we identify four distinct regimes-linear, solitonic, turbulent, and superfluid-emerging from the interplay between pump strength, laser detuning, and injected momentum, which together control the balance between kinetic and interaction energies in the quantum fluid. The different regimes are characterized through real-space and momentum-space observables, as well as through the temporal first-order coherence function. We show that turbulence occupies a well-defined and extended region of parameter space, marked by spontaneous vortex nucleation, and a pronounced reduction of temporal coherence, providing a clear signature of nonstationary dynamics. By constructing quantitative phase diagrams, we delineate the transitions between the various regimes and identify multiple pathways connecting solitonic, turbulent, and superfluid behaviors. Finally, we demonstrate that the turbulent regime persists over experimentally realistic parameter ranges compatible with state-of-the-art GaAs-based micro-cavity platforms, establishing counter-propagating polariton flows as a robust and versatile setting for the study of driven-dissipative quantum turbulence in two dimensions.

Screened scalar fields such as the symmetron provide a viable description of dark energy yet their laboratory detection remains challenging. We propose an optomechanical scheme to constrain symmetron interactions using two optically levitated nanospheres inside a cavity. The symmetron-mediated interaction induces an effective coupling which leads to a measurable splitting in the optomechanical resonance spectrum. We forecast constraints in the regime $μ\sim 10^{-2}$eV-$10^{-4}$ eV, which shows that this approach can improve existing laboratory bounds by up to several orders of magnitude, demonstrating the sensitivity of optomechanical spectroscopy to screened fifth forces.

Magnetic fields represent a critical component of astrophysical research, laying the foundation for interpreting high-energy astrophysical activity across galactic scales. In this work, we investigate the parametrized post-Einsteinian (ppE) waveform imprints induced by the external magnetic fields of Bertotti-Robinson and Bonnor-Melvin black holes, with the aim of distinguishing such magnetic effects from environmental influences--particularly for massive black holes posited to reside at galactic centers. We first compute the ppE frequency-domain waveform for a small black hole inspiraling into a massive Kerr-Bertotti-Robinson (KBR) black hole, which corresponds to a Kerr black hole embedded in an external magnetic field. We find that the leading-order correction arising from the magnetic field is at the $-2$ post-Newtonian (PN) order relative to the quadrupole term, while the next-leading-order correction is at $-1.5$ PN, originating from the spin of the black hole. We further examine the effects of a spinning Kerr-Bonnor-Melvin (KBM) black hole, whose leading-order magnetic correction is at $-3$ PN (consistent with the preceding result), whereas its spin-induced correction is also at $-1.5$ PN. The leading-order ppE corrections for both KBR and KBM black holes do not appear degenerate with any modified theory of gravity effects; nonetheless, we demonstrate that they resemble the gravitational pull contributions from additional matter with power-law distributions of index $γ=1$ and $γ=0$, respectively. As a result, future gravitational wave (GW) observations detecting $-3$ or $-2$ PN order corrections will infer their origin as either magnetic field effects or matter environmental influences.

Quantum code surgery offers a flexible, low-overhead framework for executing logical measurements within quantum error-correcting codes. It encompasses several fault-tolerant logical computation schemes, including parallel surgery, universal adapters and fast surgery, and serves as the key primitive in extractor architectures. The efficiency of these schemes crucially depends on constructing low-overhead ancilla systems for measuring arbitrary logical operators in general quantum Low-Density Parity-Check (qLDPC) codes. In this work, we introduce a method to construct an ancilla system of qubit size $O(W \log W)$ to measure an arbitrary logical Pauli operator of weight $W$ in any qLDPC stabilizer code. This new construction immediately reduces the asymptotic overhead across various quantum code surgery schemes.

Black-hole uniqueness, i.e., the statement that all stationary vacuum black holes in the universe are described by the Kerr solution, is expected to break in theories beyond General Relativity. This breaking can take a particularly strong form, if several branches of black-hole solutions beyond the Kerr solution coexist. We find an example of a theory that exhibits such strong breaking. In this theory, a cubic coupling of a scalar field to the Gauss-Bonnet invariant triggers black-hole scalarization through a non-linear instability of the Kerr solution. At large spin, curvature-induced and spin-induced scalarization mechanisms compete at fixed sign of the coupling. This results in a rich phase structure of black-hole solutions and continuous as well as discontinuous transitions between the different branches of black holes.

For classical field theories with probabilistic initial conditions the classical field observables are an idealization. Their arbitrarily precise values poorly reflect the characteristic uncertainty in the presence of substantial fluctuations. We propose to describe this system by observables based on fluctuating fields. In terms of these "statistical observables" the probabilistic classical field theory becomes a quantum field theory. Non-commuting operators are associated to observables. The quantum rules follow from the laws for classical probabilities. We construct the functional integral for the quantum field theory, and discuss in detail the classical relativistic Klein-Gordon equation with interactions.

We experimentally validate a unified input--output model that incorporates internal and external coupling phases in a room-temperature cavity magnonic system. By explicitly accounting for phase effects, the model provides full control of interference-induced antiresonances and enables a clear interpretation of the transition from level repulsion to level attraction. Nonreciprocal transmission -- originating from internal phases -- is accurately reproduced under specific coupling conditions. Quantitative agreement between experiments and simulations is obtained across all coupling regimes, demonstrating a practical route toward phase-controlled cavity--magnon devices.

We examine the observational viability and physical implications of the Gong-Zhang (GZ) dark--energy equation-of-state parametrizations using exclusively late-time cosmological probes. Two one-dimensional parametrization models, GZ-Type~I and GZ-Type~II, are constrained with Type~Ia supernovae (Union3, Pantheon+SH0ES, and DES-SN5YR), DESI baryon acoustic oscillations, and cosmic chronometer measurements of $H(z)$. Bayesian inference combined with information-criteria diagnostics shows that both parametrizations provide competitive alternatives to $Λ$CDM, while the GZ-Type~II model is consistently favored, exhibiting reduced parameter degeneracy and stronger Jeffreys-scale support. Beyond background expansion tests, we employ configuration entropy as a thermodynamically motivated probe of structure formation. We demonstrate that the entropy-production rate sensitively traces the impact of dynamical dark energy on late-time gravitational clustering while preserving standard early-time behavior. Our results establish the Gong-Zhang framework as a physically transparent and observationally consistent extension of $Λ$CDM, with configuration entropy providing a complementary diagnostic of late-time cosmic acceleration.

We present a loop quantization of the marginally bound Lemaître-Tolman-Bondi (LTB) model, describing the gravitational collapse of pressureless dust in spherical symmetry. The full quantum LTB model is constructed as a collection of non-interacting shells, each governed by an individual single-shell loop quantum dynamics. We show that the single-shell evolution is non-singular and that wave packets initially peaked on a collapsing trajectory undergo a bounce at Planckian energy densities and subsequently follow an expanding classical trajectory, resolving the classical central curvature singularity. We also compare the loop quantum theory with the Wheeler-DeWitt quantization of the same model and assess the accuracy of the loop quantum gravity effective theory in reproducing the full quantum dynamics. Specifically, we find that initially collapsing wave packets generically develop an interference pattern at the bounce, which suppresses the accuracy of the effective theory near the center of the dust cloud.

The process tensor framework to open quantum systems provides the most general description of multi-time correlations in non-Markovian quantum dynamics. A compressed representation of a process tensor in terms of matrix product operators (MPO) can be used for numerically exact calculations of multi-time correlation functions in systems strongly coupled to a non-Markovian reservoir. We show here that the numerical scaling for computing multi-dimensional spectra can be significantly improved using a time-translation invariant MPO representation of the process tensor obtained from the uniform time-evolving matrix product operator (uniTEMPO) method. In particular, this approach provides a spectral representation of the non-Markovian dynamics that gives direct access to correlation functions in Fourier-space, avoiding explicit real-time evolution. We calculate linear and 2D electronic spectra for an example system and discuss the performance and numerical scaling of our simulations.

Determining the physically accessible unitary dynamics of a quantum system under finite Hamiltonian resources is a central problem in quantum control and Hamiltonian engineering. Dynamical Lie algebras (DLAs) provide the fundamental link between available control Hamiltonians and the resulting quantum dynamics. While the structural classification of DLAs is well-established, how to systematically engineer and reshape these algebraic structures under realistic physical constraints remains largely unexplored. In this work, building upon recent results on direct sums of identical DLAs, we develop a unified framework for engineering Hamiltonian-driven quantum dynamics based on DLAs: (i) constructing qubit-efficient direct-sum Hamiltonian structures via spectral decomposition of Hermitian operators, enabling parallel simulation of multiple quantum subsystems; (ii) identifying Hamiltonian modifications that preserve full controllability, including the $\mathfrak{su}(2^N)$ algebra, even when additional physically motivated control terms are introduced; and (iii) engineering restricted Hamiltonian sets that confine quantum dynamics to target subalgebras through irreducible Lie-algebra decompositions, providing a principled approach to symmetry-based dynamical reduction. By bridging these Lie-algebraic insights with practical control objectives, our framework provides a systematic pathway for engineering expressive and resource-efficient unitary evolutions, thus unlocking greater structural flexibility of Hamiltonian-driven quantum systems.

We construct and classify asymptotically flat, static, spherically symmetric hairy black hole solutions in $U(1)$ gauge-invariant scalar-vector-tensor (SVT) theories carrying both electric and magnetic charges. Extending previous studies beyond the quadratic sector, we systematically incorporate cubic and quartic interaction terms in the presence of the magnetic charge. We derive a consistency condition for the quartic interaction that eliminates higher-order derivative terms induced by the magnetic charge, ensuring the theory remains second-order. We classify the obtained solutions based on their symmetry properties: shift-symmetric couplings yield secondary hair governed by the Noether current, whereas $φ$-dependent interactions generate primary hair. Crucially, our analysis reveals that the magnetic charge plays a key role in activating specific interaction sectors such as the cubic coupling $\tilde{f}_3$, which does not appear in the field equations in purely electric configurations. We identify solution branches that are intrinsic to the magnetic charge, as they cease to exist in the vanishing monopole limit ($P\to 0$). Furthermore, we demonstrate that the scalar hair exhibits distinct asymptotic decay rates depending on the interaction type, suggesting possible variations in observational signatures. Finally, we verify the global regularity of these solutions by connecting analytic expansions with numerical integration.

Quantum weight reduction procedures ease the implementation of quantum codes by sparsifying them, resulting in low-weight checks and low-degree qubits. However, to date, only few quantum weight reduction methods have been explored. In this work we introduce a simple and general procedure for quantum weight reduction that achieves check weight 6 and total qubit degree 6, lower than existing procedures at the cost of a potentially larger qubit overhead. Our quantum weight reduction procedure replaces each qubit and check in an arbitrary Calderbank-Shor-Steane code with an ample patch of surface code, these patches are then joined together to form a geometrically nonlocal Layer Code. This is a quantum analog of the simple classical weight reduction procedure where each bit and check is replaced by a repetition code. Due to the simplicity of our weight reduction procedure, bounds on the weight and degree of the resulting code follow directly from the Layer Code construction and hence are easily verified by inspection. Our procedure is well suited for implementation in modular architectures that consist of surface code patches networked via long-range interconnects.

We introduce a new resource-efficient scheme for fault-tolerant quantum computation known as `macroscale multiplexing' (or simply `Macromux'), that utilizes scalable postselection to significantly improve the threshold of a given fault-tolerant protocol against both Pauli and erasure errors. Macromux is a hierarchical method for postselecting on constant-size space-time windows of a fault tolerant protocol, requiring only constant additional overheads. The method can be straightforwardly implemented for any fault-tolerant protocol and in any architecture that has access to routing and memory, such as linear-optical fusion-based architectures. We construct fault-tolerant protocols that, to our knowledge, have the highest thresholds in the literature; we perform simulations of fusion-based schemes based on the surface code, showing a maximum possible increase in Pauli thresholds of up to a factor of $\sim6$ (from $1.0\%$ to $5.9\%$). Our schemes are highly-resource efficient, and can for example, double the loss thresholds of some photonic fusion-based protocols using as little as $3 \times$ overhead.

We investigate the spatial profile of the curvature perturbation generated in multi-waterfall hybrid inflation models, which are known to produce various topological defects. Using the lattice simulation code \acl{STOLAS}, based on the stochastic formalism of inflation, we analyse six cases by varying the number of waterfall fields $n$ and the functional form of the inflaton potential (``Quadratic'' and ``Cubic'' cases). Our statistical analysis shows that the \acp{PDF} and power spectra are broadly consistent with the so-called stochastic-$δN$ algorithm. The ``Cubic'' case also exhibits a characteristic upper bound in the \ac{PDF}, as discovered in our previous work, that suppresses \acl{PBH} formation while potentially affecting halo formation. Furthermore, we employ the Euler characteristic as a topological diagnostic tool to identify the structures of the waterfall fields as well as the curvature perturbation. We find that the topological defects, such as domain walls ($n=1$), cosmic strings ($n=2$), and monopoles ($n=3$), are reconnected during inflation into finer structures by the stochastic noise, making their correlation lengths much smaller than the Hubble scale at the critical point of the waterfall phase transition counterintuitively. The Euler characteristic also implies global structures of the curvature perturbation for $n=1$, though we do not conclude if they are due to the domain wall, because neither the strings ($n=2$) nor monopoles ($n=3$) leave such structures. The global structures of the curvature perturbation will provide a novel probe for the physics of the early universe.

The effective field theory (EFT) of cosmic acceleration provides a model-independent framework for describing dark energy and modified gravity, yet many of its defining operators remain weakly constrained by existing observations. We show that measurements of horizon-scale metric fluctuations with a lunar laser interferometer can directly probe the kinetic sector of the EFT of dark energy, enabling constraints on operators governing scalar perturbation dynamics rather than only the background expansion history. In particular, we demonstrate sensitivity to the EFT kinetic coefficient $M_2^4$ and the associated sound speed of dark energy, $c_s^2$. This establishes a qualitatively new observational handle on the microphysical consistency conditions of late-time acceleration models, allowing broad regions of EFT parameter space to be probed, constrained, or potentially discovered.

We show that a driven-dissipative cavity-magnonic dimer supports multistability with coexisting symmetric and symmetry-broken steady states. The interplay between magnon Kerr nonlinearity and photon tunneling induces magnon self-trapping, leading to a persistent population imbalance between the two resonators. In the vicinity of saddle-node bifurcations, the system exhibits critical slowing down, with relaxation times far exceeding the intrinsic dissipation scale. Focusing on quan- tum correlations, we analyze the quantum fidelity and mutual information between the intercavity magnon modes. We find that both the infidelity and the mutual information increase sharply near the phase boundaries, providing clear quantum signatures of the multistable and symmetry-broken phases. Our results establish cavity magnonic dimers as a versatile platform for exploring nonlinear nonequilibrium physics in hybrid quantum systems.

We present the results of directed searches for continuous gravitational waves from unknown neutron stars in five Milky Way globular clusters. We carry out the searches in the LIGO data from the first eight months of the fourth LIGO-Virgo-KAGRA observing run using the WEAVE semi-coherent program, which sums matched-filter detection-statistic values over many time segments spanning the observation period. No gravitational wave signal is detected in the search band of 20-475 Hz for assumed source ages greater than 300 years. Injections of simulated continuous wave signals in the data indicate that we achieve the most sensitive results to date across most of the explored parameter space volume, obtaining median 95% confidence level upper limits as low as $\sim 4.2 \times 10^{-26}$ near 282 Hz for NGC 6397.

We propose a universal approach based on Hamiltonian inverse engineering to realize a set of parameterized two-qubit gates. This method possesses unique advantages to simultaneous control of transitions among four energy levels, providing a simpler and effective way to construct composite two-qubit gates with fewer operations than traditional methods. Applied to silicon double quantum dots (DQDs), one can realize a one-step fSim gate and a B gate with only one pulse switch. Of note, the method can be further integrated with various optimization theories to enhance gate performance. Based on quantum optimal control theory, we develop a high-fidelity fSim gate scheme with experimentally feasible pulse shapes, featuring an average gate time of 50 ns and a theoretical fidelity of 99.95% in the presence of decoherence and approximation error. By incorporating geometric quantum gate principles, we propose a combined geometric and dynamic fSim gate scheme. Numerical simulations demonstrate that this hybrid scheme exhibits stronger robustness against systematic errors compared to the purely dynamic approach. Our method is generalizable to arbitrary two-qubit physical systems, offering a feasible pathway for rapidly and robustly constructing composite two-qubit gates.

There has been increasing interest in developing efficient quantum algorithms for hard classical problems. The Network Signal Coordination (NSC) problem is one such problem known to be NP complete. We implement Grover's search algorithm to solve the NSC problem to provide quadratic speedup. We further extend the algorithm to a Robust NSC formulation and analyse its complexity under both constant and polynomial-precision robustness parameters. The Robust NSC problem determines whether there exists a fraction (alpha) of solutions space that will lead to system delays less than a maximum threshold (K). The key contributions of this work are (1) development of a quantum algorithm for the NSC problem, and (2) a quantum algorithm for the Robust NSC problem whose iteration count is O(1/sqrt(alpha)), independent of the search space size, and (3) an extension to polynomial-precision robustness where alpha = alpha_o/p(N) decays polynomially with network size, retaining a quadratic quantum speedup. We demonstrate its implementation through simulation and on an actual quantum computer.

Continuous-variable-discrete-variable (CV-DV) quantum simulators offer a natural route to simulating bosonic dynamics relevant to many branches of physics and chemistry. However, programmable simulation of arbitrary dynamics is an outstanding challenge. In particular, simulating anharmonic dynamics, which is ubiquitous across the physical sciences, is challenging due to the highly harmonic nature of oscillators used in CV-DV simulators. Here, we experimentally demonstrate programmable CV-DV quantum simulation of anharmonic dynamics in a range of double-well potentials, implemented in a trapped-ion system. We synthesise the time-evolution operators using a bosonic-quantum-signal-processing subroutine, which allows the potential to be tuned between experiments by controlling classical experimental parameters. We observe coherent dynamics in various double-well potentials, where a wavepacket tunnels through the potential barrier, and we suppress this effect by programmatically introducing asymmetry.

In this paper, we investigate the evolution of the early universe within the framework of fractional Newtonian cosmology. By constructing a suitable fractional potential, we show that the cosmological evolution can naturally originate from a non-singular pre-inflationary regime. We find a natural transition time, separating the pre-inflationary and inflationary regimes, characterized by the balance of the corresponding forces. By analyzing the dynamics near the transition time, we show that the inflationary phase emerges as a stable dynamical attractor. We show that the fractional force vanishes and undergoes a sign change at a point very close to the end of inflation. We then determine the small separation between the force zero point and the end of inflation, and show that it leads to a meaningful relation between the number of $e$-folds and the fractional parameter $α$, ensuring consistency with observations and resolving the horizon problem. Moreover, our results demonstrate the existence of a graceful exit from inflation, followed by an exact radiation-dominated solution with the standard time dependence and an $α$-dependent normalization.

We study ultra-Planckian $2\to2$ scattering in an Abelian gauge theory coupled to agravity, the scale-free and renormalizable realization of quadratic quantum gravity. Focusing on charged fermions and scalars interacting with the photon and the higher-derivative graviton, we present compact analytic expressions for the unpolarized squared matrix elements for a broad set of tree-level processes, including photon--photon, fermion--fermion, fermion--photon, scalar--fermion, scalar--photon, scalar--scalar, and annihilation channels. In contrast to purely graviton-mediated analyses, we retain systematically the photon--graviton interference contributions and verify explicitly the independence of the results on the gravitational gauge-fixing parameter. The amplitudes display characteristic forward/backward enhancements associated with small momentum transfer, amplified by the $1/p^{4}$ graviton propagator, while their high-energy scaling reflects the underlying dimensionless gravitational couplings. Moreover, for all channels analyzed the corresponding differential cross sections exhibit the universal ultra-Planckian scaling $dσ/dΩ\propto 1/s$, where $s$ is the Mandelstam invariant (the squared center-of-momentum energy). Our results furnish a unified amplitude-level description of how higher-derivative gravity reshapes familiar QED scattering at ultra-Planckian energies and provide analytic building blocks for further studies of IR definitions and UV consistency in agravity with matter.

Warped compactifications of massive type IIA supergravity on a 3-sphere with spacetime-filling O6/D6 sources are known to admit a half-maximal gauged supergravity description in 7D. We study the effect of introducing open string degrees of freedom (scalars and fluxes) in such dimensional reductions, associated with the spacetime-filling sources. From the 7D supergravity point of view, this can be realized by coupling the gravity multiplet with extra vector multiplets and adding new components to the embedding tensor describing the gauging. The scalar potential of the underlying theory exhibits novel AdS7 vacuum solutions, with and without supersymmetry. Finally, we explore the net of domain wall solutions interpolating between the different pairs of vacua, and present analytical as well as numerical solutions.

Indirect searches for dark matter using dwarf spheroidal galaxies are limited by systematic uncertainties in modeling diffuse gamma-ray backgrounds. We present a likelihood-based comparison of locally constructed empirical background models and theoretically-motivated models that incorporate the Fermi-LAT diffuse background. The empirical models we study include both an independent-binning approach and a covariance-based approach that captures cross-energy correlations. Using ensembles of blank-sky regions and information criteria which account for model complexity, we find that empirical background descriptions provide a statistically competitive fit to gamma-ray data on degree scales in high-latitude regions.

We describe our method for rewriting dimensionally regulated Feynman parameter integrals in the Minkowski regime as a sum of real, positive integrands multiplied by complex prefactors. This representation eliminates the need for a contour deformation, which is one of the main computational bottlenecks in numerical integration. We demonstrate clearly how the method works on two examples, and benchmark the performance against contour deformation as implemented in pySecDec, where we observe performance gains of up to several orders of magnitude. We describe an improvement in the resolution procedure using the Generic Cylindrical Algebraic Decomposition algorithm, which generalises our method to any Feynman integral, including those with massive propagators.

Ultra-high energy cosmic rays impinge on the atmosphere and induce air shower cascades, in which huge numbers of particles are produced. By traveling through the Earth's atmosphere and magnetic field, these particles create a noticeable effect on the electric field at the surface. In this article, we demonstrate that parameters of the shower longitudinal development can be inferred from mapping the time dependence of the observed electric field to the emitted electric field as a function of slant depth along the shower axis.

We investigate the hierarchical flavor structure of the Standard Model in a Froggatt-Nielsen (FN) framework, where the spontaneous breaking of a $U(1)_{\rm FN}$ symmetry by a complex flavon field generates fermion masses and mixing patterns through higher-dimensional operators. Extending the setup with three right-handed neutrinos (RHNs), light neutrino masses arise via the Type-I seesaw mechanism. Allowing complex FN coefficients enables a consistent description of the CKM and PMNS matrices while inducing CP-violating signatures in meson decays. Building on our previous work, where the lightest RHN acts as a viable dark matter (DM) candidate produced through freeze-in or freeze-out mechanisms, we investigate the origin of the baryon asymmetry of the Universe. The heavier RHNs generate a lepton asymmetry through out-of-equilibrium decays, including both Standard Model channels and additional flavon-induced processes in which the flavon appears as an intermediate or final-state particle. We compute the corresponding one-loop CP asymmetries and incorporate these effects in the Boltzmann equations. We show that although freeze-in and freeze-out DM production occur in two qualitatively distinct regions of the FN symmetry-breaking scale $v_φ$, successful thermal leptogenesis can be achieved in both regimes. In the large-$v_φ$ (freeze-in-compatible) region the results approach the standard leptogenesis limit, while in the freeze-out-compatible region the lower value of $v_φ$ implies lighter RHNs, requiring resonant enhancement. This tightly constrained framework, in which $v_φ$ simultaneously controls RHN masses and the interaction strengths of the flavon and DM sectors, provides a predictive and unified description of flavor hierarchies, neutrino masses, CP violation, dark matter, and baryogenesis within a single effective theory.

We analyse an inverse seesaw scenario with 3+3 gauge singlets. The flavour structure is determined by a flavour symmetry, Delta (3 n^2) or Delta (6 n^2), n integer, and CP and their residual groups among charged leptons and the neutral states. For the latter, the Dirac mass matrix of the gauge singlets carries all non-trivial flavour structure. Consequently, the heavy sterile states form three pseudo-Dirac pairs which have in general distinct masses. We discuss the signal strength of different charged lepton flavour violating processes. Ensuring that the lepton mixing angles can be accommodated at the 3 sigma level or better, we find that the current bounds on the branching ratios of mu -> e gamma, mu -> 3 e, tau -> l gamma and tau -> 3 l, l=e, mu, as well as the rate of mu-e conversion in nuclei do not strongly constrain the considered parameter space, while the limits expected from the upcoming experiments Mu3E, COMET and Mu2e will have a relevant impact.

We investigate the production of millicharged particles (mCPs) by the Schwinger mechanism in thunderstorms in the atmospheres of different planets in the Solar system. We study both fermion and scalar mCPs; for scalar mCPs, we take into account the effect of Bose enhancement. We use the observational data of planetary atmospheres obtained by satellite missions to establish constraints on the charge and mass of mCP particles. The best constraints came from the observation of thunderstorms in Saturn's atmosphere: $q > 10^{-11}$ for fermionic mCP and $q > 10^{-24}$ for bosonic mCPs. These constraints are, to the best of our knowledge, the best in the literature.

We revisit the idea of numerically integrating the differential form of Feynman integrals. With a novel approach for the treatment of branch cuts, we develop an integrator capable of evaluating a basis of master integrals in double and quadruple precision, with significantly smaller run times than other tools. This opens the door to evaluating higher complexity Feynman integrals on the fly in Monte Carlo generators, and enables a cheaper and easy to parallelise generation of grids for the topologies with prohibitive computational times. To show its performance, we test one- and two-loop integral families, achieving evaluation times in double precision of milliseconds and hundreds of milliseconds, respectively. We comment on the results and suggest room for improvement.

We investigate the \(J/ψ\) direct production mechanism in the rare exclusive Higgs decay \(H\to J/ψ+γ\) within nonrelativistic QCD (NRQCD), which provides a clean probe for extracting the charm-quark Yukawa coupling to the Higgs boson. The Principle of Maximum Conformality (PMC) is used to remove conventional renormalization-scheme and scale ambiguities in the next-to-next-to-leading-order (N\(^2\)LO) perturbative QCD series. Large logarithmic contributions arising from Yukawa coupling renormalization are resummed, providing a reliable foundation for subsequent analyses. Using the experimentally measured leptonic decay width of \(J/ψ\) and the N\(^2\)LO perturbative result, we extract the factorization-scale-dependent long-distance matrix element \(\langle J/ψ({\bm ε})|ψ^{\dagger}{\bm σ}\cdot{\bm ε}χ(μ_Λ) |0\rangle\). Combining this with the factorization-scale-dependent short-distance coefficient, we obtain a factorization-scale-invariant decay width for the channel. Compared with earlier predictions in the literature, our fixed-order result for \(Γ(H\to J/ψ+γ)\) is more robust and precise, with good convergence and no renormalization- or factorization-scale dependence. We find \(Γ(H\to J/ψ+γ) = (6.4574^{+0.3995}_{-0.3995}) \times 10^{-11}\) GeV, where the uncertainty is the quadratic sum of contributions from \(Δα_s(m_Z) = \pm 0.0009\), \(ΔΓ_{J/ψ\to e^+e^-} = \pm 0.10\ \text{GeV}\), \(Δ\overline{m}_c(\overline{m}_c) = \pm 0.0046\ \text{GeV}\), and the estimated magnitude of N\(^3\)LO contributions from Bayesian analysis. This work demonstrates for the first time how the PMC can be applied to obtain fixed-order perturbative predictions that are invariant under both renormalization and factorization scale variations.

We present our updated results on the intrinsic width of the profile of the flux tube in (2+1)-dimensional Yang-Mills theory with SU(2) gauge group. We identify the intrinsic width as the characteristic length scale of the exponentially decaying tails of the profile of the flux tube. Inspecting a broad range of temperature, we check that this length does not depend on the length of the flux tube. Our estimations of the intrinsic width show a constant value at low temperature and a growing trend approaching the deconfinement temperature that can be understood from the universality class of the phase transition via the Svetitsky-Yaffe mapping.

We investigate the spectrum of $T_{ΥΥ}$ tetraquark candidates within a coupled-channels framework. The analysis includes all $L\leq2$ combinations of $Υ(1S)$, $Υ(2S)$, $η_b(1S)$, and $η_b(2S)$ in the $J^P = 0^\pm, 1^\pm, 2^\pm$ sectors. The meson-meson interaction is derived from an underlying constituent quark model through the resonating group method, and the properties of the states are obtained from poles of the scattering matrix. We find a rich spectrum of resonant, and virtual, states distributed between the $η_b(1S)η_b(1S)$ and $Υ(2S)Υ(2S)$ thresholds. The pattern of poles exhibits approximate heavy-quark spin symmetry multiplets. Several states are dominated by a single channel and can be associated with threshold-driven structures, while higher-mass resonances show sizable mixing among channels involving radially excited bottomonia. The predicted widths range from tens to several hundred MeV. Branching ratios indicate that many states couple predominantly to final states with at least one excited bottomonium, whereas only a subset of the spectrum is expected to be visible in the $η_b(1S)η_b(1S)$, $η_b(1S)Υ(1S)$ and $Υ(1S)Υ(1S)$ channels. These results provide quantitative guidance for experimental searches of fully heavy tetraquarks and offer a test of coupled-channel dynamics and heavy-quark spin symmetry in the $bb\bar b\bar b$ sector.

In short-baseline experiments such as LSND and MiniBooNE, an excess of electron neutrinos has been observed, originating from a muon neutrino beam. To address this anomaly, in the line of many works, we investigate various neutrino mixing schemes involving eV-scale sterile neutrinos alongside three active neutrinos. Using updated experimental and global fit data, we studied neutrinoless double beta decay for three different schemes such as 3+1, 1 + 3, and 2 + 2, which involve one sterile neutrino and three active neutrinos. We have done analysis of these schemes for normal hierarchy (NH) as well as for inverted hierarchy (IH) frameworks, and constrained the sterile neutrino mass in light of current and future neutrinoless double beta decay experiments. The 3+1 scheme is found to be the most viable and at the level of $3σ$ the mass of sterile neutrino with respect to the lightest neutrino mass ($m_{\text{lightest}}$) is restricted to $4.75~eV$ for the NH and $4.72~eV$ for the IH. Additionally, the limits on the sum of four neutrino masses are determined to be $4.81~eV$ for the normal hierarchy and $4.78~eV$ for the inverted hierarchy. The updated analysis of all these schemes would help us in understanding physics governing neutrinoless double beta decay and limit on the mass of sterile neutrinos.

Coherent elastic neutrino-nucleus scattering (CE$ν$NS) is predominantly governed by vector neutral-current interactions, with subleading contributions arising from the axial current in nuclei with non-zero ground-state spin. Experimentally, the extraction of axial-current contributions has been so far of little interest, mainly because of the challenges its measurement entail. In this work, we investigate the relative size of the vector and axial components for target materials currently employed by the neutrino and dark matter experimental communities. We identify fluorine-based compounds as the most promising targets for probing the axial-current event rate. Among them, octafluoropropane ($\text{C}_3\text{F}_8$) emerges as a particularly suitable candidate, given its widespread use in spin-dependent dark matter searches and its relevance for upcoming dedicated CE$ν$NS experiments. Considering both pion decay-at-rest and reactor neutrino fluxes, we show that such measurements can allow an indirect determination of the axial coupling at the $\sim 10\%$ level, depending on flux uncertainties and detector thresholds. We further emphasize that measurements of the axial current will allow to probe spin-dependent new physics scenarios through CE$ν$NS.

Questions at the intersection of the AdS/CFT correspondence and quantum information theory motivate the study of projectors in sequences of subalgebras of finite-dimensional commutative associative semisimple algebras $\mathcal{A}$, obtained by incrementally adjoining one generator at each step to produce a non-linear generating set for $\mathcal{A}$. We define degeneracy graphs, which are finite layered tree graphs whose nodes represent projectors in the successive subalgebras. Using combinatorial properties of the degeneracy graph, we give a simple formula for constructing a linear basis of $\mathcal{A}$ in terms of monomials in the generators.The nodes can be labelled by formal variables corresponding to the eigenvalues of the generators added at each layer. We prove that the construction is compatible with the required counting of projectors in $\mathcal{A}$, and give explicit constructions of the projectors in terms of the monomials, in the cases of one- and two-layer degeneracy graphs with arbitrary numbers of nodes. More generally, we provide extensive computational evidence for the invertibility of the matrix relating the proposed monomial basis to the projector basis, by evaluating its determinant. In the 1-layer case, this is a Vandermonde determinant. A simple formula for the non-vanishing determinant in the general layer case is conjectured and supported by the computational data. The construction is illustrated with examples including centres of symmetric group algebras and maximally commuting subalgebras generated by JucysMurphy elements. We outline applications of the monomial basis to algorithms for constructing matrix units in non-commutative semisimple algebras, with relevance to orthogonal bases of multi-matrix gauge-invariant operators and to quantum information theory.

In quark-gluon plasma (QGP), at high temperatures $T$ the spontaneous generation of color magnetic fields, $b^3(T), b^8(T) \not = 0$ (3, 8 are color indexes), and usual magnetic field $b(T) \not = 0$ happens. Also, the Polyakov loop and related to it the $A_0(T)$ condensate, which is solution to Yang-Mills imaginary time equations, create. Recently, with the new type two-loop effective potential, which generalizes the known integral representation for the Bernoulli polynomials and takes into consideration the magnetic background, these effects were derived. The corresponding effective potential $W(T, b^3, b^8, b, A_0 )$ was calculated either in SU(2) gluodynamics or full quantum chromodynamics (QCD). The values of magnetic field strengths at different temperatures were calculated and the mechanism for stabilizing the background due to $A_0(T)$ was also discovered. In present paper, we concentrate on the one-loop quark contributions. In particular, we derive the effective vertexes, which couple magnetic fields and $A_0$. The vertexes result in new specific effects signalling the creation of QGP in heavy ion collision experiments. Key words: spontaneous magnetization, high temperature, asymptotic freedom, effective potential, $A_0$ condensate, effective vertexes.

We present a combined analytical and numerical study of the effective action of field theory cosmic strings in the Abelian-Higgs model in flat space. Starting directly from the underlying solitonic field theory description, we provide a systematic derivation of the low energy effective action and present evidence for the absence of nontrivial curvature correction terms when only the translational Goldstone modes are retained. Using this framework, we extend the effective theory to include higher energy fluctuations of the soliton profile, which map to massive degrees of freedom propagating on the worldsheet. We show that the leading curvature contribution enters only through the coupling between these massive modes and the worldsheet Ricci scalar. We validate the resulting effective theory via lattice simulations of the full field theory equations of motion in flat space, implemented with Adaptive Mesh Refinement to capture the string dynamics across different scales. The numerical simulations confirm the dynamics obtained using the effective action in its validity range. Furthermore, they also demonstrate the existence of the predicted parametric instability of excited strings that drives the transfer of energy from massive excitations to the Goldstone sector.

We realize a homological block of a knot complement in $S^3$ for $G_{\mathbb{C}}=SL(2,\mathbb{C})$ as a half-index of a 3d $\mathcal{N}=2$ theory via an expression of the homological block as an inverted Habiro series by working out some examples, which we expect to extend to general knots. Also, by choosing a certain set of poles in the integral expression of the half-index, we obtain the colored Jones polynomial.

We demonstrate that TeV-scale heavy neutral leptons (HNLs) responsible for inverse-seesaw neutrino mass generation can simultaneously fix the cosmological abundance and decay properties of dark matter (DM). The spontaneous breaking of lepton number gives rise to a pseudo-Nambu-Goldstone boson that serves as a light DM candidate, whose mass originates from a small explicit breaking term. The same HNLs that generate neutrino masses produce the DM via freeze-in and mediate its decay into neutrinos, leading to a tight correlation among neutrino masses, DM relic abundance, and DM lifetime. For collider-accessible TeV-scale HNLs, the observed relic density and lifetime constraints point to sub-GeV DM, yielding observable neutrino signals at next-generation detectors such as Hyper-Kamiokande, DUNE, and JUNO. This framework establishes a predictive and experimentally testable link between neutrino mass generation and dark matter.

We address the strong CP problem: why the physical QCD angle theta-bar must be extraordinarily small given the stringent bounds on the neutron electric dipole moment. Peccei-Quinn axion models can relax theta-bar dynamically, but rely on an approximate global symmetry expected to be violated by quantum gravity and face severe astrophysical and cosmological constraints. We propose Discrete theta Projection, an axionless, gauge-protected resolution obtained by gauging a finite cyclic subgroup $Z_N $of the $2π$ shift symmetry of theta. Coupling QCD to a compact, local and gapped topological sector orbifolds the path integral, identifying theta values that differ by $2π/N$ and admitting only instanton sectors whose topological charge lies in $Z_N$. In the large four-volume limit the vacuum energy becomes the lower envelope of the orbifold images, so the theory dynamically selects the branch closest to the CP-symmetric point, enforcing $|\barθ| \le π/N$ without assuming any prior smallness. Because the discrete shift is gauged, continuous renormalization of theta is forbidden; the construction can be formulated via higher-form/two-group structure with integer-quantized couplings fixed by anomaly inflow, ensuring radiative and gravitational stability and satisfying mixed gauge-gravity consistency conditions. The framework predicts a neutron EDM suppressed by $1/N$, no axion signatures, no domain-wall/isocurvature issues, and lattice diagnostics: piecewise-analytic theta dependence with cusps at odd fractions of the reduced period and a global curvature scaling as $1/N^2$. We provide the EFT construction, a nonperturbative proof of vacuum projection, a full anomaly analysis, and UV embeddings (including discrete clockwork chains) that generate large effective N while preserving integrality and consistency throughout.

These are course notes for the 'Introduction to holography' Master level course at University of Cologne. The goal of the course is to give a pedogogical introduction to holography. Holography is a popular approach to quantum gravity, in which a theory of gravity can be described by a lower-dimensional boundary theory that itself has no gravity. The most concrete known example of a holographic model is the AdS/CFT correspondence, where the gravitational theory has a negative cosmological constant (the universe is asymptotically Anti-de Sitter) and the boundary theory is a conformal field theory. Symmetry plays a very important role in this duality. We therefore start the course with a review of Poincaré symmetry in quantum field theory, before moving on in the second chapter to conformal symmetry in conformally invariant quantum field theories or CFT's. Then we move to the basics of AdS physics in chapters 3 and 4, which will already reveal hints to the existence of a duality with CFT. After gathering the basic ingredients (CFT and AdS), in the second half of the course we are ready to formulate the AdS/CFT correspondence (chapter 5), including finite temperature AdS/CFT (chapter 6), which involves black holes and their thermodynamics in the gravitational theory (chapter 7). We end the course with an introduction to entanglement in AdS/CFT and the origin of statements that 'gravity emerges from entanglement' in holography.

There are strong arguments favoring the Flavor Democracy hypothesis (or the Democratic Mass Matrix approach) within the Standard Model framework. However, the large mass of the top quark ($m_t >> m_b, m_τ$) poses an obstacle to the functioning of Flavor Democracy in the three SM family scenario. While a fourth Standard Model generation could have provided a natural resolution, this possibility is now almost entirely excluded by precision data on Higgs boson production and decay rates. The Flavor Democracy hypothesis can be elegantly resurrected through the introduction of Vector-Like Leptons (VLLs) and Vector-Like Quarks (VLQs), which naturally accommodate the observed fermion mass hierarchies while remaining consistent with current experimental constraints. Currently, experimental searches for VLLs conducted by the ATLAS and CMS collaborations rely on a highly constrained Restricted Model. This model imposes a mass degeneracy between charged and neutral VLLs within a doublet and assumes the absence of right-handed neutrinos. Consequently, current results are valid only for the restricted model and do not cover a more realistic, general scenario. Therefore, to accurately reflect the physical reality, it is imperative to conduct a comprehensive re-evaluation that incorporates all viable decay channels.

At muon colliders, muon decays naturally produce intense electrons and positrons with unique features, namely high energies, high repetition rates, and small intrinsic uncertainties, that are unavailable at existing accelerator facilities. We quantitatively study the feasibility of extracting such particles in two representative future muon collider designs, IMCC and $μ$TRISTAN. Using Monte Carlo simulations with the corresponding design parameters, we study the spatial, angular, and energy distributions of decay electrons and positrons in the curved sections of the collider ring. We find that typical deflections of $0.1-10~\mathrm{mrad}$ can be achieved even for high-energy electrons carrying large energy fractions ($\simeq 0.6 - 1.0$) of the muon beam energy, with the ring bending magnets (or magnets providing an equivalent field) effectively serving as a pre-septum magnet, that partially deflects the beam before the main septum magnet, suggesting that the extraction scheme could be practically feasible. Exploiting the distinct beam properties of IMCC and $μ$TRISTAN, we propose complementary search strategies, missing energy and momentum searches for dark matter at $μ$TRISTAN and visible-decay searches for axion-like particles and light scalars at IMCC, which probe parameter space beyond the reach of current and other proposed experiments.

This thesis presents a comprehensive study of the renormalisation of chiral gauge theories in dimensional regularisation (DReg) at the multi-loop level. We employ the mathematically consistent Breitenlohner-Maison/`t~Hooft-Veltman (BMHV) scheme with non-anticommuting $γ_5$, whose modified algebraic relations induce a spurious violation of gauge and BRST invariance. A central focus is the systematic restoration of the broken symmetry, for which we provide a transparent and fully algorithmic procedure based on the quantum action principle. A major achievement of this work is the complete 4-loop renormalisation of an Abelian chiral gauge theory -- the highest-order application of the BMHV scheme to date. This calculation is made possible by an automated, high-performance computational framework incorporating several optimised algorithms. Our results demonstrate that a rigorous, self-consistent treatment of $γ_5$ is feasible even at very high loop orders. We further analyse dimensional ambiguities and evanescent details corresponding to different implementations of the regularisation, and identify practically efficient prescriptions for $D$-dimensional fermions and gauge interactions. Building on these insights, we present the complete 1-loop renormalisation of the full Standard Model (SM) in the BMHV scheme, providing a first step towards a fully self-consistent multi-loop renormalisation of the SM and establishing a solid foundation for future high-precision electroweak phenomenology.

The indirect determination of the total width of the detected Higgs boson that is carried out by the experimental collaborations at the LHC relies on the assumption that the coupling modifiers for the on-shell and off-shell couplings are the same. However, physics beyond the Standard Model affecting the on-shell and off-shell regions differently could invalidate this assumption, so that the actual width of the detected Higgs boson could be larger than the bounds obtained under this assumption. Relaxing the assumption and investigating different types of extensions of the Standard Model, we analyse under which conditions a larger total width of the detected Higgs boson is compatible with all experimental and theoretical constraints. For the considered scenarios of scalar extensions with an additional state contributing as a resonance or at the loop level, we find that the indirect bounds obtained by ATLAS and CMS remain valid over large parts of the parameter space, with the exception of parameter regions where the additional particles have relatively small masses. We discuss the potential of experimental searches for new particles to further constrain such scenarios. Based on the existing experimental and theoretical constraints we conclude that relaxing the assumption of equal on-shell and off-shell coupling modifiers that is used in the experimental analyses at the LHC yields an upper bound on the total width of the detected Higgs boson in realistic extensions of the Standard Model that is only weakened by up to a factor of about two compared to the case where this assumption is valid.

We present a method to determine the sensitivity of a closed dielectric haloscope to axion dark matter. Dielectric haloscopes aim to probe the theoretically well-motivated axion mass range of ~26 $\mathrmμ$eV to ~500 $\mathrmμ$eV by utilizing a stack of dielectric disks and a mirror to enhance the axion-photon conversion within an external magnetic field. Their conversion volume is nearly axion-mass independent, thereby favoring large-scale designs to increase sensitivity. The large volume causes simulations to be computationally expensive and time-consuming. This paper presents a simple model that can be used to determine the sensitivity of the experiment with minimal computational resources. The model is able to describe the electromagnetic response of a closed dielectric haloscope, accounting for realistic geometric imperfections, as well as the noise introduced by the receiver system. It is applied to data taken with a MAgnetized Disk and Mirror Axion Experiment (MADMAX) prototype within the 1.6 T Morpurgo magnet at CERN. This work underpins the first axion dark matter search using a dielectric haloscope and provides the foundation for future dark matter searches with MADMAX.

We test exact marginality of the deformation describing the resolution of a $\mathbb{Z}_2$ orbifold by analyzing the closed superstring equations of motion to third order in the size, including $α'$ corrections. We find that the third order correction is unobstructed for all deformation moduli. We are also able to reproduce the Eguchi-Hanson gravitational instanton up to the second order in the field theory limit with a suitable choice of moduli.

We present a data-driven investigation of the exhaustive ensemble of no-scale type IIB flux vacua constructed in \cite{Chauhan:2025rdj}. Using a combination of linear and non-linear dimensionality-reduction techniques, we analyse both flux and moduli spaces and demonstrate that the effective dimensionality of the underlying 12-dimensional flux space is substantially reduced. A central component of our study is a physics-informed autoencoder, which provides a non-linear compression of the flux and moduli data into a low-dimensional latent space. The learned latent representation organises vacua according to desired features and, in particular, isolates distinguished regions associated with small values of the flux superpotential $|W_0|$, revealing non-trivial correlations that are not captured by linear methods. In parallel, we apply tools from topological data analysis, specifically persistent homology, to probe the global structure of the vacuum distribution. This allows us to identify robust, long-lived topological features in both moduli and flux subspaces. This work is a necessary step for developing foundation models in string phenomenology.

Studies of QCD phase transition signals are often conducted under spatially uniform temperature conditions. However, the influence of spatial temperature gradients on the signals emerging at the phase interface in the fireball generated by heavy-ion collisions has not yet been fully explored. Based on an Ising-like effective potential, we study the locally equilibrated systems with temperature gradients. In a 2D disk geometry, the low-energy fluctuation spectrum is explicitly resolved into radial and angular momentum modes. The nonlocal correaltions of singular eigen-mode exhibits strong anisotropy, which are long-ranged along isotherms but suppressed radially due to the thermal geometry of the system. Unlike homogeneous systems where the zero-momentum mode dominates, correlations in such inhomogeneous system result from the superposition of a series of zero and non-zero angular momentum modes with comparable contributions. We extract the singular angular momentum modes and establish their connection to experimentally observable anisotropic flow. We find azimuthally sensitive observables may offer a previously unexplored avenue for detecting the QCD phase transition.

We investigate the magnetic moments of strange hidden bottom pentaquark states within the constituent quark model considering molecular and compact configurations. The system with quark content qqqbb is analyzed in three scenarios a baryon meson molecular configuration bq1bq2q3 a diquark diquark antiquark configuration bq1q2q3b and a diquark triquark configuration bq1bq2q3. The negative parity states with are studied for strangeness. We find that for the dominant spin couplings and maximally aligned configurations the diquark diquark antiquark qqqbb and diquark triquark bqqqb descriptions yield identical or numerically very close magnetic moments indicating that in the hidden bottom sector the magnetic properties are governed primarily by the global spin flavor structure rather than clustering details. A systematic suppression with increasing strangeness and a clear spin hierarchy are observed in all configurations. Due to the large bottom quark mass, heavy quark contributions are strongly suppressed, making the magnetic moments primarily sensitive to light strange spin correlations. These results provide theoretical benchmarks for future experimental studies of exotic multiquark states.

We investigate families of 3d $\mathcal{N}=2$ chiral quiver gauge theories conjectured to be dual to M2-branes probing toric SE$_7$ singularities. Geometrically, these families correspond to toric diagrams without internal points. At the field theory level, the models are constructed via an un-higgsing procedure applied to non-chiral quivers. While the moduli space of these theories was shown to match M-theory expectations, determining the $N^{3/2}$ scaling of the free energy remained an open problem for over a decade, with positive results emerging only very recently. In this work, we address this challenge by reformulating the three-sphere partition function as a hyperbolic hypergeometric integral. Using exact integral identities, we show that the free energy reduces precisely to that of non-chiral quivers with chiral flavors, for which the $N^{3/2}$ scaling is already established. Physically, this mathematical identity corresponds to the equivalence of three-sphere partition functions under a generalization of Giveon-Kutasov duality to chiral quivers. Our results thus provide a large $N$ duality between the chiral quivers and non-chiral quivers with chiral flavors, confirming the $N^{3/2}$ scaling for the chiral quivers under study.

Understanding the internal structure of the proton-including the distributions of quarks and gluons and their contributions to proton properties such as mass-remains a central challenge in quantum chromodynamics (QCD). While quark generalized parton distributions (GPDs) have been studied extensively, a unified approach that simultaneously extracts quark parton distribution functions (PDFs), gravitational form factors (GFFs), and gluon GPDs from experimental constraints is still lacking. Moreover, the role of gluons in proton mass generation, particularly through the trace anomaly mechanism, requires deeper theoretical and phenomenological exploration. In this study, we first extract quark GPDs in protons using parameterization method based on the electromagnetic form factors provided by Light-Front Holographic QCD (LFHQCD) from which we derive both quark PDFs and their GFFs. We then extend this approach to model gluon GPDs. Our calculations show consistency with experimental data and lattice QCD results and successfully reproduce soft Pomeron behavior. Furthermore, we investigate near-threshold $J/ψ$ production using gauge/string duality to quantify the contribution of the trace anomaly to the proton mass. Our results demonstrate that the parameterization method provides a consistent framework for describing both quark and gluon structure, bridging GPDs, PDFs, and GFFs. The analysis of $J/ψ$ production confirms that the trace anomaly contributes significantly ($\sim 23\%$) to the proton mass, with the calculated cross-section dependence on momentum transfer $t$ in agreement with experimental observations. This work advances the understanding of proton structure by integrating quark and gluon degrees of freedom and elucidating the origin of proton mass within QCD.

We study the influence of charm quark dynamics on the chiral phase structure of Quantum Chromodynamics (QCD) using the recently developed miniDSE scheme of the Dyson-Schwinger equations. By calculating the quark propagator in $2+1$ and $2+1+1$ flavor QCD, we quantify the impact of including the charm quark as a dynamical degree of freedom on the QCD phase diagram. Our results show that the charm quark induces a moderate but noticeable shift to lower chemical potential in the location of the critical endpoint (CEP) by approximately 2-3%. The result in this work indicates that the heavy-flavor dynamics can subtly influence the QCD phase structure and should be taken into account in particular for searching the CEP of QCD.

This MSc dissertation is based on the papers arXiv:2502.12694 and arXiv:2408.14813. The AdS/CFT correspondence provides a powerful framework for modeling strongly coupled gauge theories and, as a consequence, investigating non-perturbative phenomena in QCD. In this work, following an overview of the ideas that encapsulate the AdS/CFT correspondence, we present a self-consistent dynamical holographic QCD model within the Einstein-Maxwell-dilaton framework, derived from the coupled field equations, to study the mass spectra and melting behavior of heavy and exotic mesons at finite temperature and density. Finite temperature analyses reveal a confinement-deconfinement transition and sequential quarkonia melting. At finite density, an increase in chemical potential accelerates meson melting, with spectral functions evolving smoothly across the phase transition line. Finally, using a nonlinear Einstein-Born-Infeld-dilaton model, magnetic field effects demonstrate a shift from inverse magnetic catalysis to magnetic catalysis, highlighting the impact of spatial anisotropy on quarkonium stability.

We conducted a HI 21cm absorption study of a sample of 147 nearby (z < 0.1) low-power radio sources with $10\,\mathrm{mJy} < S_{1.4\,\mathrm{GHz}} < 30\,\mathrm{mJy}$ and $\log(P_{1.4\,\mathrm{GHz}}/\mathrm{W\,Hz^{-1}}) = 20.5-23.7$, using the Five-hundred-meter Aperture Spherical radio Telescope. By investigating the origin and kinematics of HI absorbing gas, we aim to study the interplay between the active galactic nucleus (AGN) and its surrounding interstellar medium. Our observations detect 12 new absorbers, combining results from the pilot survey (three absorbers out of 26 sources), yielding a detection rate of $\sim10.2^{+3.1}_{-2.0}\%$. The detection rate in our sample is lower than in higher-power samples, which is likely due to emission dilution and the dominance of extended sources, indicating a gas-rich and star-forming-dominated population in low-power sources. Among new detections, most line profiles are narrow and show velocities close to systemic ones, consistent with rotating disks, while four show disturbed kinematics indicative of inflows or outflows. The fraction of outflow candidates rises with radio power, while the fraction of inflow ones remains constant, suggesting the effect of radio emission on driving HI outflows. In our sample, compact sources show a higher HI detection rate than extended sources. Contrary to expectations from higher-power samples, MIR-bright sources at low-power radio do not exhibit a higher HI detection rate or more disturbed kinematics. In low-power radio sources, blueshifted absorption occurs only in Seyferts and low-ionization nuclear emitting regions, indicating the connection between atomic outflows and the ionization state of AGN.

We apply the unimpeded framework to perform a fully Bayesian reanalysis of the DESI DR2 data, using nested sampling with PolyChord to compute evidences for $Λ$CDM and seven extensions across combinations of DESI DR1/DR2, Planck CMB, supernovae (Pantheon+, Union3, DES-SN5YR, DES-Dovekie), and DES-Y1 weak lensing. The Bayesian Ockham's razor penalises extended models, yielding weaker or opposite preferences compared to $Δχ^2$-based analyses. For DESI DR2 BAO combined with Planck CMB alone, the DESI collaboration's $3.1σ$ frequentist preference for $w_0w_a$CDM is eliminated entirely: we obtain ${\ln B = -0.57{\scriptstyle\pm0.26}}$, modestly favouring $Λ$CDM. Adding the corrected DES-Dovekie supernova calibration maintains this concordance (${\ln B = -0.01{\scriptstyle\pm0.27}}$). However, when the original DES-SN5YR calibration is included instead, the DESI collaboration's $4.2σ$ result survives the Bayesian Ockham penalty as a $3.07{\scriptstyle\pm0.10}\,σ$ preference (${\ln B = +3.32{\scriptstyle\pm0.27}}$). That this signal persists despite the Ockham penalty makes the role of tension quantification essential: our analysis traced the preference to the DES-SN5YR calibration error, which introduced a $2.95{\scriptstyle\pm 0.04}\,σ$ conflict with DESI DR2 within $Λ$CDM -- a tension that stands out from the grid -- reduced to $1.96{\scriptstyle\pm 0.04}\,σ$ once the calibration was corrected. With the calibration corrected, the Bayesian evidence for dynamical dark energy vanishes.

The electron temperature is a crucial parameter for the determination of the gas-phase metallicity of galaxies. Low electron temperature is expected for metal-rich galaxies, theoretically. We report the discovery that temperature, as measured through auroral-to-strong line ratios of O$^+$, trends in reverse directions at 12+log(O/H) $\geq$ 8.7. This trend remains consistent regardless of the emission line fitting method employed and is not attributable to contamination or dust attenuation correction. Notably, this phenomenon is not observed in other low-ionization ions, such as S$^+$ and N$^+$, which also probe electron temperature. The results are verified in two independent datasets. We analyze the potential cause for the high [OII] auroral-to-strong line ratios at high metallicities, finding that no specific reason could account for that. This finding challenges the fundamental principles of the direct $T_e$ method for metallicity measurement, warranting further investigation into its physical interpretation.

The sub-parsec circumnuclear density profiles of galaxies represent a key element in our understanding of the accretion history and fuel availability of supermassive black holes (SMBHs). Observations that directly resolve sub-parsec scales in galaxies require extremely high resolution and generally hot (bright) environments, making this impossible in all but the nearest active galaxies. Transient accretion events onto previously quiescent SMBHs, such as a tidal disruption event (TDE), offer a new avenue to understand SMBHs and their environments. Radio-bright outflows from TDEs directly probe the ambient density at $10^{-3}-1$ pc scales, allowing direct constraints on the circumnuclear density of TDE host galaxies (i.e., quiescent galaxies). Here we present, using radio observations of a sample of 11 TDE hosts, a new methodology for fitting observed TDE radio emission to constrain their sub-parsec circumnuclear density profiles. Our findings reveal that TDE host galaxies exhibit circumnuclear density profiles remarkably consistent with the expectations of a simple Bondi accretion flow ($n_e\propto R^{-3/2}$). Under the assumption of a Bondi profile, we present a new method to jointly fit the outflow mass and ambient densities, in order to constrain the Bondi accretion rate and temperature. For the TDE host galaxies in our sample, we constrain a sample average Bondi accretion rate Eddington fraction of $\log_{10}f_{\rm{Edd}} = -3.96^{+0.30}_{-0.38}$ (as well as individual fits to each host). This work provides a methodology by which radio observations of TDEs can provide powerful constraints on the sub-parsec density distribution of quiescent SMBHs -- well inside the Bondi sphere. This opens up a new observational avenue to constrain sub-parsec gas distributions in a broad range of galaxies.

Thermonuclear X-ray bursts from the surface of accreting neutron stars are the most common astrophysical explosions in our galaxy. They provide a unique window into the physics of neutron stars, the physics of matter under extreme conditions, and the physics of astrophysical thermonuclear explosions. X-ray bursts are powered by a broad range of nuclear reactions that need to be understood to interpret observations. The relevant nuclei are mostly neutron deficient and unstable, and thus experimental information and theoretical understanding is limited and an active area of research in nuclear science. We review the current status of the nuclear physics of X-ray bursts, with special emphasis on new experimental and theoretical information on a large number of reaction rates. As such we provide an overview of the broad experimental and theoretical methods currently used to advance the nuclear physics of X-ray bursts. The new information is used to update the public JINA REACLIB database with 32 new reaction rates based on experimental information, and a new dataset of theoretical statistical model reaction rates where no experimental information is available. Using several models for X-ray bursts that are powered by mixed hydrogen and helium burning, we take advantage of the updated nuclear data to review the current understanding of the nuclear reaction sequences in such X-ray bursts, the modeling of light curves, and predictions of the composition of nuclear ashes.

The processes governing protostellar mass growth remain debated, although episodic accretion is now understood as a key feature of protostellar evolution across all masses. Luminosity bursts have been observed in both low- and high-mass protostars, but the overall statistics remain limited, especially for high-mass objects. Over the past decade, numerical simulations of high-mass core collapse have provided a theoretical framework for interpreting protostellar variability, yet additional observational constraints are required to determine the characteristics and importance of bursts. In this work, we analyse data from GASTON-GP programme, which mapped a 2.4 square degrees region of the Galactic plane (centred at l = 24 deg) at 1.15 and 2.00 mm using NIKA2 on the IRAM 30 m telescope. The survey obtained 11 epochs over four years, offering the first opportunity to study millimetre variability in a large sample of massive protostellar sources. From the combined dataset, we constructed catalogues of 2925 compact sources at 1.15 mm and 1713 at 2.00 mm. Using a dedicated relative calibration scheme, we generated millimetre light curves for around 200 high-signal-to-noise sources and identified one variable candidate. However, it is not protostellar. Consequently, we report no robust detections of variable protostellar sources in GASTON field. This is the direct consequence of observational limitations (i.e., sensitivity, resolution) combined with the lack of any 100-fold luminosity bursts during the observations, which is consistent with estimates inferred from isolated core collapse simulations. This study highlights the need for future high-resolution, high-cadence surveys to constrain the accretion histories of massive protostars.

The Microchannel X-ray Telescope on board the Space-based multi-band astronomical Variable Objects Monitor (SVOM) satellite detects and localizes the X-ray afterglow of gamma-ray bursts. One year after the launch, this paper presents the in-flight performance of the scientific analyses conducted by the on-board computer. After summarizing the analysis steps, the paper reviews the on-board results obtained with 15 gamma-ray burst afterglows detected by the telescope between October 2024 and August 2025. For all bursts, the localization uncertainty is estimated to be below 2 arcmin, as required by the mission design. On average, the measured position is found to be 40 arcsec away from the position measured by other experiments with a better sky resolution. Moreover, we show that the on-board analysis provides a precise sky location for the burst only a few seconds after the beginning of the observation. Taking advantage of an efficient very-high-frequency antenna network, this information is quickly collected on the ground and disseminated to other observation facilities. This low-latency strategy is critical for the multi-wavelength and multi-instrument follow-up program of SVOM.

Our understanding of the early Universe has long been limited by biased galaxy samples selected through various color criteria. With deep JWST infrared imaging, mass-complete galaxy samples can now be studied up to $z \sim 8$ for the first time. However, recent work has revealed systematic uncertainties in measuring physical properties of galaxies based solely on JWST/NIRCam and HST photometry, due to their limited wavelength coverage. This highlights the need for supplementary data, particularly in the rest-frame UV and near-infrared. Here we present the ULTIMATE-deblending project, which will eventually deliver self-consistent UV-to-Radio photometry for galaxies detected in deep JWST surveys, including both NIRCam and MIRI data. In this first paper, we release a 50-band photometric catalog spanning CFHT/U to JWST/MIRI F1800W, covering a total of 627.1 arcmin$^2$ across two JWST/PRIMER fields. We detail the reduction of JWST imaging data, the photometric procedures, and the SED-fitting methodology used to derive galaxy properties. Compared with photometry including only HST and JWST bands, the inclusion of deblended low-resolution photometry from ground-based telescopes improves the accuracy of photometric redshifts by $\sim$40%, while reducing the outlier fraction by $\sim$60%. This galaxy sample can serve as a key reference for statistical studies of galaxy formation and evolution in the early universe. All catalogs and JWST mosaics from the ULTIMATE-deblending project will be made publicly available.

The Tremaine-Weinberg (TW) method and its variations provide the most direct means to measure the pattern speeds of galactic bars. We establish a unifying framework by deriving an integral form of the continuity equation over an arbitrary closed loop. This naturally defines a local pattern speed for any chosen region in a galactic disk (including bars and spirals). We demonstrate that this intuitive formalism recovers all standard variants of the TW method as special cases corresponding to specific choices of the integration loop. In this paper, we validate this framework and demonstrate its diagnostic power. By applying it to a diverse set of test cases from the TNG50 simulation, including face-on prototype barred galaxies and highly constrained Mock Milky Way standard configurations, we show that this formalism accurately recovers both constant global pattern speeds and radially varying profiles. Rather than relying on rigid geometric approximations, our method naturally differentiates coherent solid-body rotators (bars) from spirals. Our results validate that this unified integral framework provides a robust, geometrically flexible, and practically extensible tool for decoding complex dynamics of galactic structures.

While it is well known that galaxies are composites of many emission processes, quantifying the various contributions remains challenging. In this work, we use unsupervised machine learning based clustering algorithms to evaluate the agreement between the clustering tools and astrophysical classifications, and hence quantify the fractional contributions of star formation processes and nuclear black hole activity to the total galaxy energy budget of radio sources. We perform clustering on the multiwavelength (optical, infrared (IR), and radio) active galactic nuclei (AGN) diagnostic spaces, using the data from the G09 and G23 fields from the Galaxy and Mass Assembly (GAMA) survey, Evolutionary Map of the Universe (EMU) survey, and the Wide-field Infrared Survey Explorer (WISE). We find that the statistical clustering recovers $\approx$ 90 % of the star forming galaxies (SFGs) and $\approx$ 80 % of the AGN. We define a new IR-radio AGN diagnostic scheme that identifies radio AGN from IR SFGs and AGN, corresponding to the KMeans cluster with approximately 90 % reliability. We demonstrate the superior power of radio AGN selection in higher dimensions using a three-dimensional space composed of directly observable parameters ($\rm W_1-W_2$ colour, $\rm W_2$ magnitude, and the 1.4 GHz radio flux density). This novel three dimensional diagnostic shows immense potential in radio AGN selection that is close to 90 % reliable and 90 % complete. We also publish a catalogue of radio sources in the EMU survey with associated probabilities for them to be active in the optical regime, through which we emphasise the philosophy of considering a galaxy to be composed of various fractions rather than a binary classification of SFGs and AGN.

We use various analytical techniques to study Young Stellar Objects (YSOs) in an area of approximately $10' \times 10'$ in the IRAS 18456-0223 star-forming region. We use archival optical (Gaia DR3) and infrared (2MASS, UKIDSS, Spitzer, WISE, and Herschel) data, along with our optical spectroscopy of three bright stars for this purpose. We identify 89 YSOs (80 Class II and 9 Class I) based on their infrared properties. Our multiwavelength SED fits of selected YSOs show that they have masses $\sim 0.1$--$7.2$ $M_\odot$ and are up to $4$ Myr old. Our Minimum Spanning Tree (MST) analysis shows that these YSOs, situated at around 600 pc, form clusters with radial extents of order 0.5 pc and mean surface densities of $\sim 60$ pc$^{-2}$. We compare UKIDSS and 2MASS data of the YSOs and find that some of them show variability. We construct maps based on Herschel data which reveal multiple column density peaks ($N_{\rm H_2} \sim 10^{22}$ cm$^{-2}$) embedded in cold ($T_d \sim 10$--$13$ K) filaments. Our near-infrared extinction map exhibits several high-$A_V$ peaks, some of which coincide with the sub-mm column density maxima. Using our optical spectra of three bright sources, we show that they are of A--K spectral type. One star shows the Li I 6707 Å line, indicating its youth.

As a work in progress, results from a chemical and physical analysis of molecular cores in early evolutionary stages concerning star formation are presented. Using archival data from the Atacama Large Millimeter Array (ALMA), a sample of 37 sources was investigated, from which spectra in the frequency range 330--350 GHz were extracted towards the central positions of the molecular cores. Transitions of HC$_3$N, H$^{13}$CN, and HN$^{13}$C were analysed using Gaussian fits, obtaining peak intensities, fluxes, and line widths. The column densities of each molecule and their abundances were estimated. The behaviour of these abundances with the temperature of the region was studied, observing positive correlations for H$^{13}$CN and HN$^{13}$C, and none for HC$_3$N. This study contributes to the characterisation of the initial conditions of the interstellar medium in early phases of stellar evolution.

We present preliminary results of an extensive research project aimed at describing the physical and chemical conditions of hot molecular cores (HMCs). Using millimeter continuum and spectroscopic data extracted from the Atacama Large Millimeter Array (ALMA) archive, we have estimated rotational temperatures ($\rm T_{rot}$) and column densities of $\rm{CH_{3}CN}$, $\rm{CH_{3}CCH}$, and A-- and E--$\rm CH_{3}OH$ for a sample of molecular cores. We present a thermal characterization of these cores, revealing the existence of temperature gradients within them. These cores are, in turn, embedded in large molecular clouds. Additionally, we estimated molecular abundances that were evaluated as tracers of the chemical evolution of these cores. Finally, in a pilot study aimed to link observations with simulations, some of the obtained molecular abundances are compared with predictions from the Nautilus code.

We investigate the effects of dynamical dark energy (DDE) on the growth of cosmic structure using a two-fluid model. This framework allows the dark energy equation of state to smoothly cross the phantom divide, in agreement with recent DESI results. In this effective description, DDE supports propagating perturbations that behave like sound waves. These perturbations induce a scale dependence in the growth of matter fluctuations and in halo bias, which can be exploited to test the dynamical nature of dark energy at the level of its fluctuations. For cluster-sized halos, the amplitude of the scale-dependent halo bias is comparable to that produced by massless neutrinos in $Λ$CDM. Using a Fisher forecast for a multi-tracer analysis of the power spectrum (P) and bispectrum (B) of galaxy number counts, we find that bispectrum information is essential to detect the scale dependence induced by the DDE sound mode. For a survey of volume $V\sim 10\, h^{-3}{\rm Gpc}^3$ at redshift $z=0.5 - 1$, a two-tracer P+B analysis could detect this scale dependence if the sound speeds of the dark energy fluids are in the range $c_s^2\sim 10^{-2} - 10^{-4}$. Lower sound speeds cause halos to experience a gravitational drag force through the excitations of sound waves. This effect impacts measurements of the growth rate inferred from cluster-sized halos at the 10\% level if one of the fluids has a very low sound speed $c_s^2\sim 10^{-5}$. Larger sound speeds $c_s^2 > 10^{-2}$ could be probed with optimal weighting schemes that reduce shot noise and increase the effective bias.

Galaxies grow very rapidly during the first Gyr of the Universe, mostly driven by high galaxy efficiencies, particularly relevant at $z>5$. This efficiency is related to high gas densities and/or compact gas distributions within these early galaxies. We want to understand the evolution of the size of galaxies at cosmic dawn, from $z=14$ to $z=5$ and its main drivers. We use the FirstLight database of 430 zoom-in cosmological simulations and radiative transfer calculations to generate synthetic images in seven JWST bands. We add observational effects, inspired by recent JWST deep extragalactic surveys. The size-mass relation is already in place at $z\simeq14$ and it shows a large diversity of galaxy sizes at a fixed mass. Extended (compact) galaxies tend to have higher (lower) specific star-formation rate (sSFR). The mass-dependent slope does not evolve significantly. This is driven by a complex interaction between stellar light and dust. Differential dust attenuation dims galaxy centers and it makes larger sizes, modifying the mass-size slope even in the rest-frame optical. At a fixed mass, galaxy size evolves very fast, as the normalization of the size-mass relation increases by 0.5 dex between $z\simeq14$ and $z\simeq6$, in 600 Myr. The SFR surface density increases with redshift, driven by higher sSFRs and smaller sizes at higher redshifts. Size evolution at a fixed stellar mass accelerates at cosmic dawn, driven by an increasing galaxy efficiency at $z\geq5$.

We investigate whether stochastic acceleration associated with pressure-anisotropy-driven magnetogenesis can generate a dynamically significant population of cosmic rays (CRs) prior to nonlinear structure formation. As magnetic fields amplify in the early Universe, the associated increase in gyrofrequency enhances pitch-angle scattering, potentially shortening the stochastic acceleration time. We derive an analytic criterion for efficient cosmological acceleration by comparing the acceleration timescale with the Hubble time, which defines a critical magnetic field and a corresponding CR turn-on redshift $z_{\rm on}$. For representative parameters, we find $z_{\rm on}\sim1.7$. To quantify the resulting particle population, we solve a Fokker-Planck equation for the isotropic proton distribution in the redshift interval $z=10\rightarrow z_{\rm on}$. Throughout most of this epoch, adiabatic expansion dominates over stochastic energization and the distribution remains close to a cooling Maxwellian. However, as the system approaches the turn-on epoch, the stochastic acceleration time decreases, allowing a mild suprathermal tail to develop. Even under optimistic assumptions corresponding to the strong-scattering limit, the maximum attainable proton energy reaches at most $\mathcal{O}(10^2)\,\mathrm{GeV}$. These results indicate that efficient CR production in the intergalactic medium is intrinsically tied to the onset of structure-formation shocks, while earlier microinstability-driven stochastic processes can provide at most a modest pre-acceleration.

We present an example cluster follow-up study with Euclid. Our target, a $z\sim 1.74$ candidate cluster nicknamed the `Puddle', was initially discovered by the Massive and Distant Clusters of WISE Survey 2 (MaDCoWS2) as a $z_{phot}\sim 1.65$ candidate cluster. It was also detected independently as a $z_{phot}\sim 1.5$ candidate with both cluster-finding algorithms in Euclid Quick Release 1 (Q1). A Keck MOSFIRE spectrum shows the brightest nucleus is at $z=1.74$ and is AGN-dominated. We focus our analysis on the galaxy population and the Brightest Cluster Galaxy (BCG), using a combination of Euclid and ancillary photometry. Compared to similar fields, we measure an overdensity of $110\pm 14$ galaxies with $H_\mathrm{E}\leq 22.25$ in a 2' radius around the BCG. We estimate that $18\pm 4$% of the completeness-corrected galaxy population is red, which is consistent with some clusters at $z>1.5$ but lower than others. \textit{Euclid} imaging reveals that six or seven galaxies appear to be assembling to form the future BCG. Spectral energy distribution (SED) fitting suggests that the merging BCG has a stellar mass of $5.7\pm 0.3\times 10^{11}\,M_\odot$ and experienced a short burst of star formation about $300\,$Myr ago. Its morphology, stellar mass, and star-formation history suggest that the proto-BCG is a more evolved version of the merging core of SPT2349$-$56. These systems indicate that multiobject mergers might be a common BCG formation process. Assuming a similar density of mergers in the Euclid Wide Survey, we expect that Euclid will discover approximately 400 assembling BCGs by the end of its mission.

We revisit the determination of the age of the Universe from galactic globular clusters, extending previous analyses by explicitly accounting for the presence of multiple stellar populations within each cluster. Using high--quality \textit{Hubble Space Telescope} color--magnitude diagrams for 69 globular clusters, we relax the standard single--population assumption, and model two stellar populations with independent ages, metallicities, helium abundances, and population fractions. The inference is performed using the full color--magnitude diagram morphology, an explicit treatment of field contamination, and a hierarchical framework that propagates non--Gaussian age posteriors. Allowing for multiple stellar populations has a negligible impact on globular cluster age estimates. The ages of the oldest populations remain fully consistent with those obtained under the single--population assumption, with differences at the $0.6σ$ level. Restricting to the metal--poor subsample ([Fe/H] $< -1.5$), we infer a dominant old component with mean age $t_{\rm GC}=13.61\pm0.25\,\mathrm{(stat)}\,\pm0.23 \mathrm{(sys)}\,\mathrm{Gyr}$. Adopting a conservative delay between the Big Bang and the formation of the first globular clusters, we obtain an age of the Universe of $t_{\rm U}=13.81\pm0.25\,\mathrm{(stat)}\,\pm0.23 \mathrm{(sys)}\,\mathrm{Gyr}$. In addition to age constraints, our analysis yields simultaneous measurements of metallicity and helium content for the different populations, including constraints on helium enrichment and population fractions which are consistent with independent determinations from the literature. These results demonstrate that globular--cluster--based cosmic chronometry is robust to stellar population complexity, reinforcing its role as a precise and largely cosmological model--independent probe of the age of the Universe.

Pulsar halos provide a unique probe of cosmic-ray propagation in the vicinity of pulsars and have important implications for our understanding of particle diffusion in the interstellar medium. However, the number of firmly identified pulsar halos remains limited. One of the main challenges is the difficulty in unambiguously confirming halo candidates through precise morphological measurements with current $γ$-ray observations. In this work, we investigate the prospects for identifying pulsar halo candidates through morphological discrimination using simulations of two advanced $γ$-ray experiments: LHAASO-KM2A and the Cherenkov Telescope Array (CTA). Using mock observations with realistic instrumental responses, we assess the ability of each experiment to distinguish diffusion-based halo morphologies from alternative simplified spatial models. Our results show that both increased photon statistics and improved angular resolution significantly enhance the power of morphological discrimination. In particular, CTA benefits from its superior angular resolution, while LHAASO-KM2A gains sensitivity from its large effective area at the highest energies. These results indicate that future $γ$-ray observations have the potential to expand the sample of pulsar halos and provide further insights into cosmic-ray transport around pulsars.

If massive primordial black holes (PBHs) exist and constitute a fraction of the dark matter, they can dramatically catalyze galaxy formation. By acting as pre-existing, high-density seeds, they can shorten the galaxy assembly time to as little as 100 Myr for up to 10^8 solar mass PBH seeds, allowing for the rapid formation of host halos. Furthermore, low surface brightness or diffuse galaxies may represent a natural outcome of this process, perhaps as the residue of halos seeded by smaller PBHs that failed to accrete a major baryonic component.

We investigate the interaction between relativistic jets and supernova (SN) ejecta as a potential origin of X-ray knots in radio galaxies, employing knot A in M 87 as a test case. By modeling the dynamical evolution of the interaction, we evaluate this scenario based on particle acceleration efficiency and spatial morphology. Our modeling indicates that the ejecta shock expands to only ~ 30 pc, which is inconsistent with the observed spatial scale of knot A (~ 60 pc). In contrast, the jet shock can successfully reproduce the observed scale after approximately 3000 yr, with the ejecta being accelerated to a bulk velocity of β~ 0.43. We fit the multi-wavelength spectral energy distribution (SED) using a one-zone leptonic framework, attributing the X-rays to synchrotron radiation from electrons accelerated up to~1 PeV at the jet shock. The derived magnetic field is approximately 70 uG in the SN ejecta rest frame, which is significantly below the equipartition value. Protons may be accelerated up to ~ EeV, supporting the hypothesis that the jets of radio galaxies (RGs) may be the potential site for ultra-high-energy cosmic-ray (UHECR) acceleration within the framework of the jet-ejecta interaction.

We performed numerical simulations along with radiative transfer calculations to reproduce an intriguing asymmetric shoulder feature in the dust-continuum emission of the protostellar disk around one of the eDisk targets, the Class 0 protostar IRAS 16544-1604 in CB 68. This is our first attempt to bridge the theoretical works of protostellar disk evolution and the eDisk observations. We found that while our hydrodynamic simulations form spiral structures caused by gravitational instability, they become less discernible after the disk is inclined and convolved with the telescope beam. The widths of the spiral structure as obtained by our numerical simulations are ~0.1-0.8 times the eDisk beam size of 4.5 au. Our modeling effor implies that the apparent absence of spiral features in the eDisk observations does not necessarily indicate the real absence of internal substructures and gravitational instability. We also found that the asymmetric shoulder structure of the continuum profile along the major axis appears when the disk is massive enough with a Toomre parameter Q~1. This mechanism offers a potential explanation for the observed, asymmetric shoulder features in the disks surrounding IRAS 16544-1604 and the other eDisk sources.

The rapid growth of imaging and spectroscopic surveys has intensified the need for efficient tools that support visual inspection, a practice that remains essential for tasks such as classification, catalog refinement, and validation of automated methods. Existing solutions, however, often require the use of multiple platforms and complex workflows to integrate heterogeneous data. To address this challenge, we present the first release of the AstroInspect (https://astroinspect.github.io), a web-based system which ensures seamless access to several astronomical resources. The system provides an intuitive graphical user interface (GUI) through which users can upload catalogs of objects defined by celestial coordinates. AstroInspect automatically enriches these catalogs with complementary information, including imaging, spectroscopic, and photometric data retrieved in real time from surveys such as the Sloan Digital Sky Survey (SDSS), the Legacy Surveys (LS), and the Southern Photometric Local Universe Survey (S-PLUS). As an example of its scientific utility, we used AstroInspect to identify H$α$ emission-line galaxies within a 7 deg radius in the direction of the Hydra I cluster (also known as Abell 1060) by visual inspection. Using a candidate set of 981 galaxies selected from S-PLUS photometric data, we produced a catalog of 80 galaxies with confirmed H$α$ emission. These results highlight the potential of AstroInspect to support efficient visual inspection workflows.

We have combined Gaia astrometry with Gaia-ESO Survey radial velocities to measure the 3D rotation of the open cluster NGC 2516. We compiled a sample of 430 members with astrometry and spectroscopy and use these to determine a distance to the cluster of 406.3 +/- 0.8 pc, which we then use to infer the 3D positions and velocities of all stars in the cluster using a Bayesian model. We identify the axis of maximum cluster rotation and measure a median rotational velocity of 0.12 +/- 0.02 km/s. We find the axis of maximum cluster rotation to be 74 +/- 17 degrees to the plane of our galaxy. We compare this rotation rate to measurements of cluster rotation in other open clusters and find that it is inconsistent with the expected dependences on cluster age and mass.