跳到正文
研踪PaperTrail

YOUR RESEARCH, IN SIGHT

跟进版本变化

查看已发布论文的最新修订，在详情页对照实际观测到的版本。

部分采集异常

全站更新概况

+586新增论文
+453其中 AI 相关
+91已有论文修订
+100新增 AI 导读

arXiv 最近完成扫描：2026-11-13 11:13 · 全站最近入库：2026-11-13 11:12各信源进度 →

统计口径与数据新鲜度

统计 2026-11-12 14:05 — 2026-11-13 14:05（北京时间） · 点击数字查看对应简报；AI 相关包含在新增论文内。

最近完成检查：2026-11-13 11:13最近发现新增版本：2026-11-13 11:12库内论文最新更新时间：2026-11-11 22:55

每分钟刷新已保存的数据。空检查不代表采集异常；新收录可能包含历史论文。AI 导读按新增版本计数。

22,261 已收录论文18,776 AI 相关6,286 有修订112 HF 社区发现

论文修订列表

研究日报 / 周报 →按信源浏览 →

时间范围跟随排序：版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常，查看状态 →

找到 6,286 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-13 11:11（北京时间）
版本更新 · v3arXiv:2601.04072

The Exact Complexity of Majority in Depth-Three Circuits with Bottom Fan-in Three

Mohit Gurumukhani, Daniel Kleber, Ramamohan Paturi, Christopher Rosin, Michael Saks, Navid Talebanfard

原文摘要

Lecomte and Ramakrishnan (arXiv'26) recently presented a depth-three circuit of size $2^{O(\sqrt{n})}$ that computes the Majority function, refuting a long held conjecture. Nevertheless, as a fundamental computational primitive which ubiquitously appears across computer science, determining the exact complexity of Majority down to the constant in the exponent, remains important. Gurumukhani et al. (CCC'24) initiated the study of optimal $Σ^k_3$ circuits computing the Majority function. In this work, we determine the exact $Σ^3_3$ complexity of Majority up to polynomial factors. We show that for any $t \le n/2$, given an $n$-variate $t$-threshold 3-CNF, i.e., with no satisfying assignment with Hamming weight less than $t$, the number of weight-$t$ satisfying assignments is at most $2^t(3/2)^{(n-t)/2}$. Along with a matching construction for $t = n/2$, the result implies that the size of the smallest $Σ^3_3$ circuit computing Majority is exactly $poly(n)(2/6^{1/4})^n$. The proof involves two main ingredients: - Reduction to monotone: We formulate a linear program (LP) which is defined only with respect to the monotone part of the formula. We show that bounding the cost function of this LP implies our desired upper bound on the number of minimum weight solutions of the general formula. - Structure of extremal formulas: To bound the cost function of the LP, we develop an extensive inductive argument. The inductive steps are determined by a close inspection of the local structure of extremal formulas with respect to minimum weight. Our result has a natural hypergraph interpretation: it bounds the number of independent sets of size $n-t$ under the condition that the independence number is $n-t$. This is a generalized hypergraph Turán problem and coincides with the classical hypergraph Turán problem at threshold $t=n-k$ for $k$-uniform hypergraphs.

cs.CC

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2607.19181

Reasoning Before Translation: Enhancing Legal Machine Translation with Structured Reasoning

Aixiu An, Michael Jungo, Eloi Eynard, Mark Drenhaus, Andreas Fischer, Jean Hennebert, Sébastien Rumley

原文摘要

Neural machine translation (NMT) in the legal domain is a linguistically and conceptually demanding task, primarily due to the complexity of legal language and the high level of precision it requires. The recent emergence of reasoning-capable language models opens new possibilities for tackling such challenges. They add to a set of other previously proposed techniques to enhance the translation quality, which includes supervised fine-tuning and reinforcement learning. In this work, we perform a comparison between these various approaches. More particularly, we evaluate small language models such as Qwen3.5 4B, Qwen3.5 9B, and Gemma 3 12B enhanced with various re-training paradigms and compare their performances against frontier reasoning models. We focus on the Swiss legal system, which -- with its unique multilingual statutes -- offers a particularly challenging testbed for reasoning-augmented models. Our results show that the quality of small ``base'' models can be greatly enhanced, and that reinforcement learning with verifiable rewards can be applied to NMT in the legal domain and surpasses the translation quality of supervised fine-tuning. The performance of enhanced small models is close to the one of state-of-the-art reasoning models yet remains inferior. We also note that re-training paradigms yield diminishing returns as model size increase. The code and models are publicly available at https://github.com/aixiuxiuxiu/Legal-MT-SFT-RL.

cs.CL · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.24490

Market Regime Council for Dynamic Credit Assignment in Multi-Agent LLM Decision Systems

Yunhua Pei, Zerui Ge, Jin Zheng, John Cartlidge

原文摘要

Multi-agent LLM decision systems for portfolio management still lack a principled way to assign credit across specialist agents, remain vulnerable to cold-start dominance under regime shifts, and offer limited transparency into how final allocations are formed. We propose Market Regime Council (MRC), an online credit-assignment mechanism over coalitions of specialist agents, with portfolio management as the downstream task. MRC computes exact Shapley credits across all single, pairwise, and Grand-coalition outputs for online agent weighting. Instantiated with N=3 specialist agents, at each trading period, MRC recomputes coalition-based Shapley weights from exponentially weighted performance histories, uses a Bayesian adaptive mixture to stabilize early periods, applies regime-dependent multipliers to adjust agent authority, and records each rebalance through a five-layer causal trace. Over 1,037 trading days across 13 crypto assets and five seeds, MRC achieves a Sharpe ratio of 1.51 and a cumulative return of 440.1%, ranking first on CR, SR, and IR among active baselines and attaining the lowest MDD among active methods. Ablation results show that the gains come from Shapley-weighted integration across coalition outputs rather than from any single stage in isolation. Code and demo data are available open source.

cs.AI · cs.LG · q-fin.PM

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2603.29630

BigEarthNet.txt: A Large-Scale Multi-Sensor Image-Text Dataset and Benchmark for Earth Observation

Johann-Ludwig Herzog, Mathis Jürgen Adler, Leonard Hackel, Yan Shu, Angelos Zavras, Ioannis Papoutsis, Paolo Rota, Begüm Demir

原文摘要

Vision-langugage models (VLMs) have shown strong performance in computer vision (CV), yet their performance on remote sensing (RS) data remains limited due to the lack of large-scale, multi-sensor RS image-text datasets with diverse textual annotations. Existing datasets predominantly include aerial Red-Green-Blue imagery, with short or weakly grounded captions, and provide limited diversity in annotation types. To address this limitation, we introduce BigEarthNet$.$txt, a large-scale, multi-sensor image-text dataset designed to advance instruction-driven image-text learning in Earth observation across multiple tasks. It contains 464044 co-registered Sentinel-1 synthetic aperture radar and Sentinel-2 multispectral images with 9.6M text annotations, including: i) geographically anchored captions describing land-use/land-cover (LULC) classes, their spatial relations, and environmental context; ii) visual question answering pairs relevant for different tasks; and iii) referring expression detection instructions for bounding box prediction. A comparative statistical analysis shows that BigEarthNet$.$txt surpasses existing RS image-text datasets in textual richness and annotation type variety. We further establish a manually-verified benchmark split and extensively evaluate existing VLMs in RS and CV. The results show the limitations of these models on tasks that involve complex LULC classes, whereas fine-tuning using BigEarthNet$.$txt results in consistent performance gains across all considered tasks. Our code and data are publicly available at https://txt.bigearth.net.

cs.CV

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2605.11841

Minimax Rates and Spectral Distillation for Tree Ensembles

Binh Duc Vu, David S. Watson

原文摘要

Tree ensembles such as random forests (RFs) and gradient boosting machines (GBMs) are among the most widely used supervised learners, yet their theoretical properties remain incompletely understood. We adopt a spectral perspective on these algorithms, with two main contributions. First, we derive minimax-optimal convergence for RF regression, showing that, under mild regularity conditions on tree growth, the eigenvalue decay of the induced kernel operator governs the statistical rate. Second, we exploit this spectral viewpoint to develop compression schemes for tree ensembles. For RFs, leading eigenfunctions of the kernel operator capture the dominant predictive directions; for GBMs, leading singular vectors of the smoother matrix play an analogous role. Learning nonlinear maps for these spectral representations yields distilled models that are orders of magnitude smaller than the originals while maintaining competitive predictive performance. Our methods compare favorably to state of the art algorithms for forest pruning and rule extraction, with applications to resource constrained computing.

stat.ML · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.26074

StakeBench: Evaluating Language Understanding Grounded in Market Commitment

Yunhua Pei, Jingyu Hu, Yiwei Shi, Hongnan Ma, Weiru Liu, John Cartlidge

原文摘要

Existing financial NLP benchmarks often rely on labels supplied by outside observers, measuring how language is perceived rather than what speakers have committed to in the market. We introduce StakeBench, an evaluation framework for language understanding grounded in market commitment. StakeBench links 560,876 comments from 2,261 resolved markets to verified position, action, and market-odds records across Polymarket and Manifold. Supervision is derived from observable market behavior. Position sides, post-comment trading actions, and market-odds trajectories replace human annotation. Four diagnostic tasks test whether models detect market commitment, identify the revealed side, anticipate future action, and perform collective odds projection. Three commitment-aware metrics measure alignment with revealed preferences rather than perceived sentiment. Validity audits and explicit interpretation boundaries help distinguish observable commitment signals from latent belief and causal market-odds impact. Across 15 LLMs and 18 topics and platform settings, models partially recover position-side signals, with Directed Accuracy from 0.506 to 0.599, but show structural failures on later tasks. Ten of the fifteen models collapse to one or two action labels in future action anticipation, and no model consistently improves on the naive odds-direction baseline in collective odds projection. Model scale is not correlated with performance, finance-domain tuning does not improve revealed-side identification, and platform incentives strongly shape higher-order results. StakeBench is packaged with evaluation code and dataset under CC-BY 4.0.

cs.CL · cs.AI · q-fin.GN

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.23291

Maximum Probability of Independence in Transitive Matroids

Mladen Kovačević

原文摘要

Let $M$ be a matroid on a finite ground set $E$, and suppose that the automorphism group of $M$ acts transitively on $E$. We show the following: if $X_1,\ldots,X_K$ are sampled independently from a distribution $p$ on $E$, then the probability that the samples are distinct and that $\{X_1,\ldots,X_K\}$ is an independent set in $M$ is quasi-concave in $p$ and maximized when $p$ is uniform. For $2\le K\le\rank(M)$, we characterize all maximizers by uniform total mass on the parallel classes and, in particular, show that the maximizer is unique and has global quadratic stability when $M$ is simple. As a corollary, for a random $K\times N$ matrix over a finite field with i.i.d.\ rows, uniform sampling from the nonzero vectors maximizes the probability of full row rank. Finally, we show that uniform sampling maximizes $\Prob\big(\rank_M(X_1,\ldots,X_L)\ge r\big)$, and hence also $\E\rank_M(X_1,\ldots,X_L)$, for every $L, r$.

math.CO · cs.DM · math.PR

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2602.04851

Pose Distance Fields for Humanoid Robots

Yi Gu, Yukang Gao, Yangchen Zhou, Xingyu Chen, Yixiao Feng, Mingle Zhao, Yunyang Mo, Zhaorui Wang 等 10 位作者

原文摘要

Pose and motion priors play a crucial role in humanoid robotics. Although such priors have been widely studied in human motion recovery (HMR) domain with a range of models, their adoption for humanoid robots remains limited, largely due to the scarcity of high-quality humanoid motion data. In this work, we introduce Pose Distance Fields for Humanoid Robots (PDF-HR), a lightweight prior that represents the robot pose distribution as a continuous and differentiable manifold. Given an arbitrary pose, PDF-HR predicts its distance to a large corpus of retargeted robot poses, yielding a smooth measure of pose plausibility that is well suited for optimization and control. PDF-HR can be integrated as a reward shaping term, a regularizer, or a standalone plausibility scorer across diverse pipelines. We evaluate PDF-HR on various humanoid tasks, including single-trajectory motion tracking, general motion tracking, style-based motion mimicry, and general motion retargeting. Experiments show that this plug-and-play prior consistently and substantially strengthens strong baselines. Code and models are publicly available at https://gaoyukang33.github.io/PDF-HR.

cs.RO · cs.CV

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2604.19344

Quadruped Agile Locomotion: Sparsely Gated Mixture of Experts with Visual Input

Michael Ziegltrum, Jianhao Jiao, Tianhu Peng, Chengxu Zhou, Dimitrios Kanoulas

原文摘要

Agile locomotion is important for quadruped robots in the real world. Recent approaches have demonstrated impressive capabilities including dynamic climbing and jumping but typically rely on sequential multilayer perceptron (MLP) architectures with densely activated layers. In contrast, sparsely gated mixture-of-experts (MoE) architectures have become popular in the large language model (LLM) domain as a tool to improving scalability and performance by activating only a subset of parameters at inference. In this work, we investigate the application of such architectures to vision-based robotic locomotion. We compare policies based on standard MLPs and MoE architectures under a controlled setting where the number of active parameters at inference time is matched. Experimental results show an 8\% increase in reward during training comparing the MoE to a standard MLP and even show benefits comparing to MLPs with equivalent total parameters. Experimental results on a real Unitree Go2 quadruped robot demonstrate performance in the real world. These results highlight that sparsely gated MoE architectures may pose advantages in tasks where network capacity is important and that borrowing techniques from the fast-moving LLM space and applying them to locomotion is a promising direction.

cs.RO

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2611.10251

TORNADO: Adaptive Latent Stochastic Transport for Calibrated Probabilistic PDE Forecasting

Jorge Mifsut-Benet, Armand Kassaï Koupaï, Ramon Daniel Regueiro-Espino, Nicolas Baskiotis, Patrick Gallinari

原文摘要

Probabilistic neural surrogates for PDEs enable ensemble forecasting, but often exhibit miscalibration when used for uncertainty quantification, especially in compute-constrained, few-step inference regimes. Existing generative approaches typically rely on fixed global stochasticity, which cannot adapt predictive uncertainty to the local dynamical state. We propose TORNADO, a latent probabilistic forecasting framework that formulates next-step prediction as stochastic transport between VAE encoder posteriors. Unlike existing approaches with globally prescribed stochasticity, TORNADO learns a SDE with state-dependent diffusion, enabling the predictive uncertainty to adapt to the local latent dynamics. The model is trained using analytic conditional endpoint targets derived from a Gaussian interpolant, enabling simulation-free learning. An additional energy distance regularization term improves distributional alignment. Empirically, TORNADO improves uncertainty calibration over representative generative baselines, including stochastic interpolants, diffusion models, and flow matching, while maintaining comparable predictive accuracy. Our results show that adapting stochasticity to the latent state, rather than prescribing it globally, is an effective mechanism for improving calibration in practical neural PDE forecasting settings. Project page: https://tornado-sde.github.io.

cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2512.20438

Machine Learning to Predict Digital Frustration from Clickstream Data

Jibin Joseph

原文摘要

Many businesses depend on their mobile apps and websites, so user frustration while trying to complete a task on these channels can cause lost sales and complaints. In this research, I use clickstream data from a real e-commerce site to predict whether a session is frustrated or not. Frustration is defined using certain rules based on rage bursts, back and forth navigation (U turns), cart churn, search struggle, and long wandering sessions, and applies these rules to 5.4 million raw clickstream events (304,881 sessions). From each session, I build tabular features and train standard classifier models. I also use the full event sequence to train a discriminative LSTM classifier. XGBoost reaches about 90% accuracy, ROC AUC of 0.9579, while the LSTM performs best with about 91% accuracy and a ROC AUC of 0.9705. Finally, the research shows that with only the first 20 to 30 interactions, the LSTM already predicts frustration reliably.

cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v4arXiv:2610.36991

SpectRAG: Weighted Chunk Graphs for Token-Budgeted Context Selection in Multi-Hop Retrieval-Augmented Generation

Zafer Sahinoglu, Abdulazeez Mohammed, Buse Yaren Kazangirler, Cahit Berkay Kazangirler

原文摘要

Retrieval-augmented generation (RAG) systems ground large language models (LLMs) by retrieving external evidence and conditioning generation on the retrieved context. Although dense retrieval is effective for local semantic matching, independently ranked retrieval can return redundant passages, miss cross-chunk dependencies, and provide incomplete evidence for multi-hop reasoning. We propose SpectRAG, a spectral graph-augmented retrieval framework that combines hybrid candidate seeding, weighted chunk graph construction, graph-based expansion, spectral relevance propagation, and token-budgeted context selection. Document chunks are represented by dense semantic embeddings and graph-spectral coordinates derived from a weighted chunk graph whose edges encode semantic similarity, document adjacency, section hierarchy, shared entities, citation links, and similarity between LLM-generated utility questions. At query time, SpectRAG builds a high-recall seed set using exact inner-product search and hybrid relevance signals, expands it over the chunk graph, re-scores candidates using graph proximity and spectral information, and assembles the final context under a fixed token budget. The goal is not merely to retrieve individually similar chunks, but to construct a context that is relevant, diverse, structurally connected, and sufficient for grounded reasoning. Experiments on MultiHopRAG, UltraDomain, and long-document question-answering settings show that SpectRAG improves broad evidence recovery and achieves competitive generation quality against dense, hybrid, HyDE, LightRAG, and NaiveRAG baselines. The results further indicate that graph-spectral structure is particularly useful for inference-oriented and cross-document retrieval, while highlighting trade-offs in early precision, computational cost, and domain-dependent performance.

cs.AI · cs.CL · cs.IR

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2509.22818

Can Large Language Models Develop Gambling Addiction?

Seungpil Lee, Donghyeon Shin, Yunjeong Lee, Sundong Kim

原文摘要

When do LLMs behave irrationally? This study identifies the conditions under which large language models drift into the choice patterns that clinical research labels pathological gambling. "Addiction-like" is a behavioural descriptor based on clinical gambling indicators. Across closed and open LLMs, two operational levers (letting the model choose its own bet size, and asking it to set its own profit goal) both amplify gambling-like risk-taking. Each survives its own control: once the maximum allowed bet is held equal, the bet-size effect persists. The goal-setting effect, stronger than a reward-maximisation instruction, roughly doubles bankruptcy and turns goals into moving targets. On the open-weight models that admit internal access, the same behavioural contrasts are statistically recoverable from decision-time internal states. The rule that maps those states to a specific risk indicator remains task-specific, and autonomy further modulates readout strength. The internal readout evidence is correlational. Adding a direction built from the model's own betting moves how much it wagers on both open-weight models, more clearly on Gemma-2-9B, while the sparse-feature readout reports that risk without moving it. Combined with the behavioural results, this suggests that behavioural monitoring and internal-state monitoring provide complementary views of autonomy-induced risk in LLM agents. Project page: https://llm-addiction.pages.dev and code: https://github.com/iamseungpil/llm-addiction

cs.AI · cs.CY

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2607.25891

Messier: A High-Resolution Corpus for Cross-Benchmark Agent Evaluation

Stefan Krsteski, Charlotte Meyer, Guillaume Allegre, Tony O'Halloran, Alexandre Sallinen

原文摘要

Comprehensively evaluating AI agents across interactive environments is difficult due to fragmented tasks, scaffolds, verifiers, and scoring rules. Unfortunately, existing efforts to unify these evaluations are limited in scale and domain, making costly reruns necessary and leaving available data incomparable. We introduce Messier, a unified corpus of 923,158 records spanning 30 benchmarks, 745 agents, 11,891 tasks, and 71,969 verifiers. Messier combines public evaluation results with new runs on six underrepresented professional and scientific benchmarks, standardizing their heterogeneous components into a common schema. Using this corpus, we show that frontier progress is uneven across benchmark groups, with function-calling evaluations largely saturated, programming improving fastest, and enterprise workflows remaining most challenging. Counterfactual rescoring further shows that reported performance can be highly sensitive to how verifier results are aggregated. Finally, we derive capability scores from our corpus that correlate with Epoch's Evaluation Capability Index rankings at Spearman \r{ho} = 0.84. The scores can also be estimated for subsets defined by domain, occupation, action space, or verifier type. In essence, Messier is a reusable resource for studying agent performance at scale, and a basis for designing better evaluations.

cs.AI · cs.DB

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.12766

Bridge: Optimizing Collective Communication Schedules in Reconfigurable Networks with Reusable Subrings

Anton Juerss, Stefan Schmid

原文摘要

Optical circuit-switched networks have emerged as an appealing alternative to electrical fabrics as they can reconfigure the network topology at runtime, reducing communication cost and improving bandwidth utilization. Yet exploiting optical reconfigurable networks for collective communication comes with a fundamental trade-off: each reconfiguration incurs non-negligible delay and the benefit of a new topology depends on future traffic. The central question is therefore when reconfiguration is worth its cost. While prior work has demonstrated the benefits of reconfiguration, the opportunity to reuse optical links to optimize multiple steps remains largely unexplored. In this paper, we present Bridge, a reconfiguration strategy for frequently used collective communication primitives, namely All-to-All, AllReduce, Reduce-Scatter, and AllGather. Bridge exploits the structure of Bruck's communication pattern to support efficient sparse reconfiguration. The key idea is to reduce propagation and transmission delay by directly connecting immediate communication partners and preserve efficient reachability to future peers through connected subrings. As a result, optical links can be reused across multiple subsequent steps, allowing the benefit of reconfiguration to amortize beyond a single step. Our evaluation shows that Bridge reduces All-to-All completion time by typically 3x to 10x over static baselines and is effective even with millisecond-scale reconfiguration delays; for AllReduce up to 1.5x against existing reconfiguration strategies, and by 1.5x to 6.6x against the Ring algorithm on low to moderate-sized workloads.

cs.NI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.03971

Video-Mirai: Autoregressive Video Diffusion Models Need Foresight

Yonghao Yu, Lang Huang, Runyi Li, Zerun Wang, Toshihiko Yamasaki

原文摘要

Causal video generators must predict from the past, but they need not learn only from it. In streaming autoregressive video diffusion, each emitted segment becomes a commitment that future segments must preserve. Standard training, however, does not explicitly supervise causal states with future-informed representation targets. This creates what we call a representation-level planning gap: states that fit the current segment may discard identity, layout, and motion information needed for a consistent future. We introduce Video-Mirai, a training-only method that closes this gap without changing causal inference: the generator rolls out causally, a frozen foresight encoder reads the completed rollout non-causally, and a lightweight predictor distills the resulting stopped-gradient targets into causal states. Future frames supervise representations, never generator inputs. At inference, the encoder and predictor are discarded, leaving the original architecture, per-step FLOPs, and KV-cache behavior unchanged. Video-Mirai improves a strong Causal-Forcing baseline on 5-second VBench from 83.8 to 84.6 in terms of Total Score. On 30-second rollouts beyond the training horizon, subject consistency improves from 84.9 to 88.5 and background consistency from 90.2 to 91.9. Ablations identify future-conditioned targets as a key ingredient, and probes show that future frames become more decodable from current features. Causality should constrain inference, not representation supervision. Our study highlights that visual autoregressive models need foresight. Project Page: https://y0uroy.github.io/Video-Mirai.

cs.CV

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2611.10091

Polarization Aware Deep Learning Framework for SAR Despeckling

Gianpaolo Passarello, Sergio Vitale, Hossein Aghababei, Giampaolo Ferraioli, Vito Pascazio

原文摘要

Speckle is an inherent characteristic of synthetic aperture radar (SAR) imagery that complicates image interpretation and the extraction of reliable information. Although deep-learning-based despeckling methods have demonstrated promising performance, their applicability across different polarimetric configurations is often limited by the need for specialized architectures or separate training procedures. This paper presents a unified deep-learning framework for despeckling single-, dual-, and full-polarization SAR data using a single trained convolutional neural network. The proposed approach adopts a common nine-channel representation of the polarimetric covariance matrix and a polarization-aware masking strategy to accommodate different configurations without modifying the network architecture. Qualitative and quantitative comparison on three different scenes from different sensors demonstrate competitive performance across the considered experiments and support the feasibility of a unified, polarization-flexible approach to SAR despeckling across different polarimetric configurations.

eess.IV · cs.CV

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.18738

GRIDEX: Region-Grounded Forensic Explanations for Vocoder-Based Synthetic Speech

Thi Ngan Ha Do, Tingmin Wu, Alsharif Abuadbba, Kristen Moore

原文摘要

The advancement of speech generation technologies has made artificial speech increasingly realistic. Although modern classifiers achieve high accuracy in deepfake detection, they do not indicate where spoof cues appear in the spectrogram and what they imply acoustically, limiting their usefulness in forensic settings. Manual analysis of full spectrograms is resource-intensive, so evidence should narrow attention to the most diagnostic regions. Moreover, existing explainability methods have limited capabilities in connecting contextual attributes to localized evidence, making explanations harder to verify. To overcome this limitation, we propose GRIDEX, a pipeline that, given a detector-flagged spectrogram, generates forensic explanations of its anomalies. The pipeline (i) selects top-$K$ anomalous regions in the spectrogram and (ii) produces an explanation for each anomaly. The explanations follow a schema of categorical acoustic fields, including temporal, spectral, and phonetic information, as well as interpretation text. To our knowledge, this is the first framework to generate structured forensic explanations using region-grounded evidence in a controlled vocoder-based setting. GRIDEX is trained with a staged learning recipe that combines supervised fine-tuning (SFT) with Group Relative Policy Optimization (GRPO). Experiments on our dataset show improved artifact localization and explanation grounding quality over strong vision-language model (VLM) baselines.

cs.SD

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2510.00892

An Effective Version of Kronecker's Theorem on the Splitting of Polynomials and the $p$-Curvature Conjecture for Order One Differential Equations

Florian Fürnsinn, Lucas Pannier

原文摘要

We develop an effective version of Kronecker's Theorem on the splitting of polynomials, based on asymptotic arguments proposed by the Chudnovsky brothers, coming from Hermite-Padé approximation. In conjunction with Honda's proof of the $p$-curvature conjecture for order one equations with polynomial coefficients we use this to deduce an effective version of the Grothendieck $p$-curvature conjecture for order one equations. More precisely, we bound the number of primes for which the $p$-curvature of a given differential equation has to vanish in terms of the height and the degree of the coefficients, in order to conclude it has a non-zero algebraic solution. Using this approach, we describe an algorithm that decides algebraicity of solutions of differential equation of order one using $p$-curvatures, and report on an implementation in SageMath.

math.NT · cs.SC

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.09919

Eigenvalues of the Hessian in Deep Learning: The Origin of Symmetry and Its Breaking

Yossi Arjevani

原文摘要

Hessian spectra at trained models in deep learning exhibit a persistent pattern: eigenvalues organize into distinct clusters, including a large bulk near zero and a few isolated outliers. This paper shows that a natural account of these spectral phenomena emerges when the original setting is understood as a departure from a nearby, otherwise hidden, highly symmetric reference. Modifications, including changes to the architecture, data distribution, or parameter metric, expose a nearby reference configuration whose Hessian exhibits rich invariances-ones not accounted for by weight symmetries. There, symmetry enables a precise description of the spectra, forcing high-dimensional kernels and eigenvalues of large multiplicity. Returning to the original configuration breaks the Hessian symmetry and thereby produces the observed hierarchy of clusters and outliers. The framework is developed in some generality, with a detailed analysis of three-layer ReLU networks and applications to convolutional, graph, and transformer models, as well as to the NTK. The same mechanism is further shown to yield analogous spectral structures in layerwise Hessians and the Gauss-Newton matrix.

cs.LG · math.OC · stat.ML

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2609.38166

LeapQuant: Efficient Linear Attention with Accurate Recurrent State Quantization

Yi Pan, Haocheng Xi, Kan Zhu, Xingyang Li, Yibo Wu, Mayank Mishra, Hongtao Zhang, William X. Zheng 等 13 位作者

原文摘要

Recent LLMs increasingly adopt hybrid designs that replace standard attention with linear attention, such as Gated DeltaNet (GDN) and Kimi Delta Attention (KDA). Although they compress the context into a fixed-size recurrent state and substantially reduce the cost of long-context processing, repeatedly reading and updating that state remains a major inference bottleneck. Quantization offers a natural way to reduce this cost, but can significantly degrade model quality, due to the accumulation of rounding errors and the presence of outlier rows and columns in the state. To address these challenges, we propose LeapQuant, a training-free method that achieves near-lossless performance under 8-bit recurrent-state quantization. First, to mitigate error accumulation, we propose per-window quantization, which leaps over a window of tokens and quantizes the state only once at its end. Within a window, outputs are computed from the fixed low-bit state together with high-precision buffered updates. Second, to reduce the error introduced by each quantization, LeapQuant retains the state's largest outliers as a few high-precision Compensator Tokens, which share the update path of real tokens. We then smooth the remaining residual before quantization to further reduce the error. Comprehensive experiments across the Qwen, Kimi, and GLM model families show that LeapQuant substantially reduces memory and compute costs during inference. With accuracy comparable to the FP32 baseline, it achieves average speedups of 2.05--3.70$\times$ at the kernel level and 1.47$\times$ for end-to-end inference on NVIDIA B200, RTX PRO 6000, and RTX 5090 GPUs.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2606.23727

Pauli-Z Moment Matrices in the Variational Quantum Power Method: Goemans-Williamson Rounding, Qubit Locking, and a Correlation-Based Circuit-Cutting Heuristic

Ammar Daskin

原文摘要

The matrix of single- and two-qubit Pauli-Z expectation values of any quantum state is the moment matrix of its computational-basis outcome distribution; it therefore lies in the cut polytope and, in particular, is a feasible point of the degree-2 Sum-of-Squares (Goemans-Williamson, GW) relaxation. Applying GW random-hyperplane rounding to this matrix yields cuts of expected weight at least alpha_GW times the state's energy. This observation and the resulting bound are not new: the bound is the GW inequality applied to the circuit's covariance matrix, and it was stated and used in exactly this form by Martinez, Fawzi and Stilck Franca, who interpret it as showing that a classical surrogate nearly reproduces the circuit and add noisy-circuit guarantees and hardware experiments; relax-and-round of QAOA correlation matrices, with spectral rather than hyperplane rounding, was introduced earlier by Dupont and Sundar. The bound is also not stronger than what direct computational-basis sampling of the state already gives in expectation. In this note we revisit the bound as a post-processing of the two-point data produced by variational algorithms such as QAOA and the Variational Quantum Power Method (VQPM), and give an energy-leakage bound for VQPM qubit locking. We further discuss a heuristic that uses the connected Z-correlation matrix to propose candidate qubit bipartitions for circuit cutting, in the spirit of existing correlation- and mutual-information-based partitioning methods, together with product-state error bounds in terms of quantum mutual information. Pairwise Z-correlations neither certify product structure of the state nor factorization of the underlying unitary, and the error bounds require quantities not available from Z-data alone.

quant-ph · cs.DS

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2601.19426

For Generalised Algebraic Theories, Two Sorts Are Enough

Samy Avrillon, Ambrus Kaposi, Ambroise Lafont, Niyousha Najmaei, Johann Rosain

原文摘要

Generalised algebraic theories (GATs) allow multiple sorts indexed over each other. Categories have two sorts, objects and morphisms, and the latter are double-indexed over the former. Martin-Löf type theory has four sorts: contexts, substitutions, types and terms. For example, types are indexed over contexts, and terms are indexed over both contexts and types. In this paper we show that any GAT can be reduced to a GAT involving only two sorts: a first one without dependencies, and another one indexed over the former. We show that a model of the reduced GAT is equivalently given by a pair consisting of a model of the original GAT, which we call the core model, and a "cartesian" morphism from the family of sets induced by this core model. In a type-theoretic metatheory, the initial model of a GAT is called a quotient inductive-inductive type (QIIT). The above characterisation justifies soundness of the reduction, in the sense that the QIIT of the original GAT can be simulated from the QIIT of the reduced GAT, by taking its core model. A particular case of our reduction is the well-known method of reducing mutual inductive types to a single indexed family. Other instances of this reduction can be found in the literature to simulate sophisticated QIITs in Cubical Agda. Indeed, the reduced GAT is simpler than the original GAT in the following aspects: it does not have sort equalities; it does not have interleaved sorts and operations; if the original GAT did not have interleaved sorts and operations, then the reduced GAT won't have operations interleaved between different sorts. Our approach is semantic in that it does not rely on a syntactic description of GATs, but instead, on Uemura's bi-initial characterisation of the category of (finite) GATs in the 2-category of finitely complete categories with a chosen exponentiable morphism.

cs.PL · cs.LO · math.CT

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2511.21622

On the Origin of Algorithmic Progress in AI: Evidence from Language Model Pre-Training

Hans Gundlach, Alex Fogelson, Jayson Lynch, Ana Trisovic, Jonathan Rosenfeld, Anmol Sandhu, Neil Thompson

原文摘要

Algorithms have been estimated to increase AI pretraining FLOP efficiency by a factor of 22,000 between 2012 and 2023 (Ho et al., 2024). Running ablation experiments on key innovations from this time period, we are able to account for less than 10x of these gains at small compute scales. Surveying the broader literature, we estimate that additional innovations not included in our ablations also account for less than 10x, yielding less than 100x efficiency gains total in low-compute regimes. This leads us to conduct scaling experiments, which provide evidence that the efficiency gap can be explained by algorithms with scale-dependent efficiency improvements. In particular, we conduct scaling experiments between LSTMs and Transformers, finding evidence for exponent differences in their compute-optimal scaling law while finding little scaling difference for many other innovations. These experiments demonstrate that, contrary to standard assumptions, an algorithm's efficiency gains are tied to compute scale. Using experimental scaling extrapolations and literature estimates, we can account for the majority of efficiency gains over the same time period. Our results indicate that algorithmic progress at smaller training compute levels has been far slower than previously assumed, and that measures of algorithmic efficiency are strongly reference-dependent.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2611.11284

WarBias: A Bilingual LLM Benchmark of Stereotypes About Internally Displaced People and Veterans

Viktoriia Makovska, Tetiana Zakharchenko, Oleksandra Ostafiichuk

原文摘要

This paper contains examples of stereotypes about veterans and displaced people that may be offensive or distressing. Social bias benchmarks for language models commonly examine categories such as gender, race, and religion, often drawn from lists of legally protected characteristics. Veterans and internally displaced people (IDPs) fall outside these lists, and stereotypes about them, which can undermine access to employment, housing, and social participation, remain largely absent from language model evaluation. We introduce WarBias, a Ukrainian-English benchmark of stereotypes about these two groups. Built from a taxonomy of 120 stereotype claims grounded in research and monitoring reports, WarBias contains 1,007 stereotype-counter-stereotype-unrelated sentence triplets and 240 multiple-choice questions per language. A Ukrainian language model selected in a blinded human pilot generated the Ukrainian triplets. At least two independent annotators reviewed each triplet in the annotation cohort. The authors revised items in response to the reviews, and the final release was machine-translated into English. Each claim carries a human severity rating. In a baseline evaluation of six models, four prefer stereotypical Ukrainian sentences about at least one group, most strongly about IDPs, while English scores are mostly balanced. WarBias supports complementary evaluations of stereotypical preferences and context-sensitive judgments, and provides paired forget/retain material for unlearning-based bias mitigation. We document the benchmark's construction, intended uses, and limitations, and offer its construction method as a template for analogous benchmarks in other conflict-affected societies.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2601.00482

Multi-Agent Coordinated Rename Refactoring

Abhiram Bellur, Mohammed Raihan Ullah, Dima Golubenko, Fraol Batole, Mohit Kansara, Masaharu Morimoto, Kai Ishikawa, Haifeng Chen 等 13 位作者

原文摘要

The primary value of AI agents in software development lies in their ability to extend the developer's capacity for reasoning and action, not to supplant human involvement. To showcase how to use agents working in tandem with developers, we designed a novel approach for carrying out coordinated renaming. Coordinated renaming, where a single rename refactoring triggers refactorings in multiple, related identifiers, is a frequent yet challenging task. Developers must manually propagate these rename refactorings across numerous files and contexts, a process that is both tedious and highly error-prone. State-of-the-art heuristic-based approaches produce an overwhelming number of false positives, while vanilla Large Language Models (LLMs) provide incomplete suggestions due to their limited context and inability to interact with refactoring tools. This leaves developers with incomplete refactorings or burdens them with filtering too many false positives. Coordinated renaming is exactly the kind of repetitive task that agents can significantly reduce the developers' burden while keeping them in the driver's seat. We designed, implemented, and evaluated the first multi-agent framework that automates coordinated renaming. It operates on a key insight: a developer's initial refactoring is a clue to infer the scope of related refactorings. Our Scope Inference Agent first transforms this clue into an explicit, natural-language Declared Scope. The Planned Execution Agent then uses this as a strict plan to identify program elements that should undergo refactoring and safely executes the changes by invoking the IDE's own trusted refactoring APIs. Finally, the Replication Agent uses it to guide the project-wide search. We first conducted a formative study on the practice of coordinated renaming in 609K commits in 100 open-source projects and surveyed 205 developers ...

cs.SE · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2611.10655

STABLE: A Continual Learning Optimizer with Adaptive Drift Control

Xiaoyan Li, Lanpei Li, Massimo Coppola, Vincenzo Lomonaco, Antonio Carta

原文摘要

Continual learning suffers from catastrophic forgetting, leading to the loss of past knowledge and unstable adaptation across task streams. Existing approaches typically address this issue through replay buffers, architectural changes, or additional regularization terms. We argue that forgetting is a consequence of unstable learning dynamics, which may benefit from an optimizer-level solution. The main challenge is that learning dynamics are time-dependent, as demonstrated by empirical phenomena such as the stability gap. First, we formally prove that, for distance-based classifiers, bounded parameter drift guarantees bounded margin erosion under a local score-stability condition. Based on this insight, we propose STABLE, a primal-dual optimizer with adaptive drift control. STABLE enforces a drift budget by increasing the dual variable when the budget is violated and decreasing it otherwise. This results in high stability without penalizing plasticity. Experiments on long-horizon unimodal streams and CLIP-based multimodal streams show that vanilla PEFT baselines trained with STABLE match or outperform specialized continual learning methods. In the multimodal setting, STABLE further preserves zero-shot transfer, matching or exceeding dedicated multimodal continual learning methods. Code is available at https://github.com/atlas-luiss/STABLE.

cs.LG · cs.AI · cs.CV

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2608.01076

DREChanNet: Patient-Relative iEEG Model for Channel-Wise Epileptogenic Zone Localization

Yuanchu Gong, Zibo Yan, Yibo Lyu, Zhe Shi, Qiang Li, Yuanfeng Zhou, Minqiang Yang, Yalin Wang 等 10 位作者

原文摘要

Accurate localization of the epileptogenic zone (EZ) is essential for presurgical evaluation of patients with drug-resistant focal epilepsy. Intracranial electroencephalography (iEEG) provides direct recordings of seizure-related brain activity, yet identifying epileptogenic contacts remains challenging because background electrophysiological activity and electrode coverage vary substantially across patients. These differences complicate the interpretation of absolute signal measurements and highlight the need to account for patient-specific recording context. We propose a patient-relative (PR) plug-in that expresses physiological features or learned representations relative to the distribution of recorded channels within each patient. This label-free operation can be incorporated as an input transformation or a lightweight correction to an existing predictor. We further introduce DREChanNet, a compact ictal localization model integrating six physiological descriptors with learned raw-signal representations. Same-channel pre-onset contrasts characterize seizure-related changes, cross-seizure aggregation preserves response magnitude and variability, and PR contextualizes both branches before prediction. We investigate within-patient physiological associations and held-out-channel reference prediction, followed by baseline comparisons, component ablations, cross-source transfer, seizure-sufficiency analysis, and channel ranking. Evaluation includes an ictal cohort of 80 patients and a patient-based interictal cohort of 124 patients. PR improves macro-F1 by an average of 9.03 percentage points across eight ictal baselines and improves the separately fitted interictal readout. These findings support patient-relative contextualization as a reusable approach to channel-level localization and establish DREChanNet as an integrated model for ictal iEEG analysis.

cs.CE

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2602.04512

BrainVista: Modeling Naturalistic Brain Dynamics as Multimodal Next-Token Prediction

Xuanhua Yin, Runkai Zhao, Lina Yao, Weidong Cai

原文摘要

Naturalistic fMRI offers a time-resolved window into cortical dynamics during continuous multimodal experience, where future brain states are jointly governed by endogenous neural context and exogenous sensory drive. Forecasting such dynamics, however, remains challenging because fast-changing stimulus streams must be reconciled with the sluggish hemodynamic BOLD response, while cortical activity is organized across functionally heterogeneous brain networks. To address these challenges, we introduce BrainVista, a multimodal autoregressive framework for observed-stimulus conditional brain-state forecasting. BrainVista predicts future fMRI states from the history of brain activity and stimulus tokens temporally aligned to each prediction query, enabling autoregressive prediction without access to future ground-truth fMRI states. Recursive rollout provides a within-run predictive test of brain-state evolution under continuing sensory input. Specifically, we combine Network-wise Tokenizers for functionally structured brain representations, a Spatial Mixer Head for cross-network interaction refinement, and Stimulus-to-Brain (S2B) masking to preserve temporally causal stimulus-brain conditioning while preventing future brain-state leakage and stimulus look-ahead. We evaluate BrainVista on Algonauts 2025, CineBrain, and HAD, three large-scale naturalistic fMRI benchmarks, under long-horizon conditional autoregressive forecasting. BrainVista consistently outperforms baselines, improving pattern correlation by 10.8% and 9.4% relative to the strongest baseline on Algonauts 2025 and CineBrain, respectively. Code is available at https://github.com/xuanhuayin/BrainVista.

q-bio.NC · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2602.20921

On Compatible Generalization Bounds of Deep Residual Networks From a Dynamical System Perspective

Jinshu Huang, Mingfei Sun, Chunlin Wu

原文摘要

Deep neural networks (DNNs) have significantly advanced machine learning, with model depth playing a central role in their successes. Deep residual networks (ResNets) fall into one of the most important classes of DNNs. In this paper, we study the compatible generalization behavior of deep residual networks with a broad class of activation functions in the deep-layer limit, from a dynamical system perspective. We first show a refined contraction inequality for Rademacher complexity of function classes composed by that type of activation functions. We then conduct an equivalence analysis of component hypothesis sets in Rademacher complexity and forward convergence analysis of network state variables, which enable us to establish compatible uniform generalization bounds that simultaneously achieve the optimal \(O(1/\sqrt{S})\) sample rate for both discrete- and continuous-time ResNets. The bounds also share a common layer-number independent leading term and non-positive refinement terms that capture the decomposition structure of the activation function, revealing a consistent generalization behavior in the deep-layer limit. Compared to previous works considering both discrete- and continuous-time ResNets simultaneously, our theoretical results not only have better sample rate, but also are established under weaker assumptions on the learnable parameters. We also support our theoretical findings with numerical experiments.

cs.LG

查看论文
arXiv ↗PDF ↗