跳到正文
研踪PaperTrail

论文库

检索计算机领域论文,按研究方向、作者或关键词筛选。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+758新增论文
+578其中 AI 相关
+94已有论文修订
+95新增 AI 导读

arXiv 最近完成扫描:2026-11-17 04:58 · 全站最近入库:2026-11-17 04:58各信源进度 →

统计口径与数据新鲜度

统计 2026-11-16 07:37 — 2026-11-17 07:37(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:2026-11-17 04:58最近发现新增版本:2026-11-17 04:58库内论文最新更新时间:2026-11-14 02:59

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。

24,106 已收录论文20,158 AI 相关7,049 有修订112 HF 社区发现

浏览论文库

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 20,158 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-17 04:57(北京时间)
版本更新 · v2arXiv:2611.08752

AttentionSwitch: Jailbreak Defense for LLMs via Joint Learning of Sparse Attention-Head Selection and Scaling

Doniyorkhon Obidov, Honggang Yu, Xiaolong Guo, Kaichen Yang

原文摘要

Large language models have achieved remarkable performance across a wide range of tasks, yet they remain susceptible to jailbreak attacks that circumvent safety mechanisms and elicit harmful or unintended outputs. Although safety can be improved through model fine-tuning, such approaches can be computationally expensive and may degrade previously acquired capabilities through catastrophic forgetting. Recent jailbreak defenses have therefore explored targeted interventions on specific model components associated with unsafe behavior. However, existing approaches typically decouple two inherently related decisions: where to intervene and how to modify the selected components. Such sequential optimization may be suboptimal because the optimal intervention location can depend on the intervention applied there, and vice versa. In this work, we introduce AttentionSwitch, the first jailbreak defense to jointly optimize intervention sites and strengths within a frozen LLM. Specifically, AttentionSwitch learns a sparse set of attention heads together with query-conditioned multiplicative scaling factors, allowing the outputs of the selected heads to be either attenuated or amplified without altering the directions of their representations. Across popular language models and jailbreak attacks, AttentionSwitch outperforms existing defenses, with particularly strong results when training data is scarce.

cs.AI · cs.CL · cs.CR

新论文 · v1arXiv:2611.17616

LLM-Guided Cluster Lifecycle Management: Deciding When and How to Evolve Behavioral Clusters

Himanshu Kulkarni, Siddharth Krishnamurthy, Jianhong Xin

原文摘要

Behavioral clusters learned over user or item interactions drift as behavior evolves: some fragment into sub-behaviors, some converge, some disappear, and new patterns emerge. Two paradigms dominate the maintenance of evolving clusterings. Threshold triggers fire a pre-specified action when a hardcoded statistic crosses a cutoff; refit-and-track (MONIC-style) re-clusters every window and infers transitions from how the new clusters align with the previous window's. Both hardwire the decision into a threshold or a re-clustering rule. We study an alternative in which a large language model (LLM) makes the lifecycle decision in place. Given a threshold-free, whole-population view of the current cluster set (statistical health, labels, member texts, residual mass), the LLM decides per cluster whether and how to act (split, merge, retire, create, or no-action) without re-clustering. Because the useful signal channel depends on how informative the labels are, we fix a richness-conditioned policy in advance: statistics when label distinctiveness is low, fused statistics+semantics when it is high. On three public datasets spanning label richness (YooChoose 0.17, MOOC 0.50, News-Category 0.70), the policy-selected LLM mode obtains the highest F1 on all three, with far fewer interventions and stronger restraint than tuned threshold heuristics and a MONIC steelman (paired Wilcoxon p <= 0.003, bootstrap 95% CIs excluding zero). The gain is concentrated more in precision than recall. The useful modality shifts from statistics to semantics/fusion as richness rises. Semantics hurt on YooChoose, fusion is roughly neutral on MOOC, and the semantic channel is required on News. One base prompt template and one fixed policy serve all three datasets, whereas the tuned baseline collapses when moved between them. We treat this shift as a practitioner diagnostic observed on three datasets, not a validated law.

cs.AI · cs.LG

新论文 · v1arXiv:2611.17610

More Agents, More Samples, or More Thinking? Untangling Coordination, Ensembling, and Test-Time Compute in Multi-Agent LLMs

Seyed Mahmoud Sajjadi Mohammadabadi

原文摘要

Evaluations of multi-agent language models often compare communicating architectures against a single greedy pass. This practice confounds coordination with sample multiplicity and inference compute. We synthesize nine primary empirical evaluations comprising 516 unique multi-agent configurations organized along a resource vector R with six dimensions and a baseline ladder with four categories. Against baselines with a single pass (Category A), multi-agent accuracy differences are negative overall and near zero when omitting the largest contributing study. When evaluated against baselines matched on sample count (Category B) or resource budget (Category C), multi-agent systems average -0.042 in accuracy, and multilevel random effects modeling on the log odds scale remains negative across sensitivity checks. Multi-agent accuracy equals or falls below the baseline in 74% of these comparable contrasts, and study means are nonpositive in almost all evaluated studies. Across studies reporting both baselines, moving from a greedy pass to a sampled baseline shifts accuracy differences by -0.038 on average. Because communication adds no information when agents share identical evidence, positive returns require partitioned evidence, external verification, or error diversity across models. We provide a reviewer checklist and accounting utilities.

cs.AI · cs.CL · cs.LG · cs.MA

版本更新 · v3arXiv:2605.22432

AMUSE: Anytime Muon with Stable Gradient Evaluation

Jueun Kim, Baekrok Shin, Jihun Yun, Beomhan Baek, Minhak Song, Chulhee Yun

原文摘要

Modern deep learning commonly relies on AdamW with prescribed learning rate schedules, but recent works challenge both components: Schedule-Free optimization removes explicit schedules via iterate averaging, and Muon improves the update geometry by orthogonalizing momentum for matrix parameters. Despite Muon's strong empirical performance, its underlying mechanism remains partially understood. We study Muon through the river-valley loss landscape, where useful training progress occurs along a flat, low-curvature bulk subspace (the river), while high-curvature dominant directions form steep valley walls that induce oscillations. We empirically show that while Muon's orthogonalization accelerates river progress by increasing the bulk component, it also amplifies dominant-direction noise, causing oscillatory trajectories. Building on this, we propose Anytime MUon with Stable gradient Evaluation (AMUSE), which integrates Muon's rapid bulk progress with the stabilizing effect of Schedule-Free averaging. AMUSE uses a time-varying interpolation coefficient that initially evaluates gradients near the fast Muon sequence for rapid adaptation, then gradually shifts toward the stable averaged sequence to suppress valley-wall oscillations. As a result, AMUSE requires no learning rate schedules and supports anytime training. Across vision tasks and large language model pretraining, AMUSE consistently improves the performance-iteration Pareto frontier over (Schedule-Free) AdamW and Muon.

cs.LG

版本更新 · v2arXiv:2410.13013

LEGAL-UQA: Testing Constitutional Question-Answering Systems for Urdu Speakers with a Parallel English-Urdu Benchmark

Faizan Faisal, Umair Yousaf, Muhammad Hassnain, Zeerak Babar

原文摘要

Legal information assistants could help Urdu speakers understand their constitutional rights, but building reliable systems requires test data and evaluation methods suited to the language. We present LEGAL-UQA, a parallel English-Urdu benchmark of 619 question-answer pairs linked to passages from Pakistan's Constitution. Each example includes the question, reference answer, and source passage in both languages. We evaluate two components separately: answering Urdu questions when the source passage is supplied, and retrieving the passage from which each question was generated. We also compare JEV, a decision model used as an automated judge, with human grades. The findings show that model rankings depend on how answer similarity is measured, while a simple word-matching method retrieves source passages more effectively than the embedding methods tested. At the evaluated cutoff, JEV agrees with human grades less often than a baseline that gives every answer full credit. An audit also identifies cases where generated references omit information supported by the source. Together, these findings highlight the need to test retrieval and answering separately, include simple retrieval baselines, and validate automated judges and reference answers with human review when engineering legal QA software for Urdu speakers.

cs.CL · cs.AI · cs.LG

新论文 · v1arXiv:2611.17608

Neural Simulation-Based Inference for Computational Cognitive Models: A Tutorial

Ti-Fen Pan, Yi Cheng, Anne GE Collins

原文摘要

Computational cognitive modeling is an increasingly essential methodology to elucidate the mechanisms underlying human and animal cognition. By fitting models to empirical data, researchers can perform parameter estimation, model comparison, and model-based neural analyses to assess competing theories, quantify individual differences, and investigate neural representations. Typically, fitting these models relies on likelihood-dependent methods (e.g., Maximum likelihood estimation or hierarchical Bayesian methods). However, for many reasonable cognitive models, computing the likelihood is computationally intractable, in which case these methods become inadequate. This substantially limits the range of theories that researchers can investigate. Simulation-based inference with neural networks (neural SBI) has emerged as a powerful toolkit to overcome this hurdle. By training deep neural networks on datasets generated through model simulations, neural SBI learns the relation between observable data and cognitive models. This approach enables all the applications of standard modeling while bypassing the likelihood computation, without constraints on empirical data size. Despite its successful adoption across diverse scientific fields, neural SBI remains underutilized in cognitive modeling, partly due to a lack of accessible, domain-tailored guidance. This tutorial addresses this gap by providing a clear workflow for applying neural SBI to cognitive models. We also provide practical use cases and recommendations for reducing the computational and simulation burden, facilitating efficient implementation in real-world research settings. By lowering the technical barriers to neural SBI, this tutorial aims to empower cognitive scientists to investigate richer theory spaces and ultimately propel scientific discovery.

cs.LG · q-bio.NC

新论文 · v1arXiv:2611.17605

Choosing Quantized Code LLMs by File Size and Energy per Solved Task

Aditi Patodiya

原文摘要

A developer running a code assistant on a laptop downloads one model file, and local inference moves the energy of every request onto that machine. A common rule says a larger model at lower precision tends to beat a smaller one at higher precision. We ran 24 GGUF files of Qwen2.5-Coder at four sizes on a fanless MacBook Air M2 and priced each in energy per solved task on 542 EvalPlus tasks. The result is a selection guide indexed by Ollama's listed file size, with an evidence label on each pick. At 1 and 2 GB it picks the q4_K_M files of the 1.5B and 3B models, both on the correctness and energy frontier under every accounting. At 3 GB as listed, the larger-model rule picks 7B-q2_K, the only measured 7B file that fits, and there it does not pay. That file brings no significant gain in solves over the 2 GB pick, 3B-q4_K_M, and the 90% CI limits any 3B-q4_K_M deficit to 2.4 points. Yet 7B-q2_K needs 2.03 times the energy per solved task on CPU, GPU and memory rails, or 1.66 to 1.81 times on whole-system power. On HumanEval+, five alternating session pairs give a ratio of 1.925 against 1.956 in the main grid. At 8 GB, 7B-q6_K buys a significant gain over the 2 GB pick at a higher energy price. From 1.5B up, q4_K_M trails the highest precision measured at its size by less than 3 points (90% CI). We release per-task results, power logs and the analysis script.

cs.SE · cs.LG

新论文 · v1arXiv:2611.17603

Offline KL-Regularized Reinforcement Learning with Stochastic Transitions

Andrew Liu

原文摘要

KL-regularized reinforcement learning is the objective behind RL fine-tuning of language models, and its data are often collected by the reference policy itself. However, beyond bandits and deterministic transitions, little is known about how much offline data it needs. In this work, we study finite discounted Markov decision processes with unknown stochastic transitions, $K$ state-action pairs and effective horizon $H$. For data that cover the regularized optimum with concentrability $C\geq 4$, every estimator needs $Ω(KCH^3/(ε\max\{ε,β\}))$ samples on some instances when $ε$ is small, where $β$ is the temperature. Under a fixed design, pessimistic soft value iteration with a variance-aware penalty attains this rate with high probability, up to logarithmic factors, when $β\geq 1.4H$. When the data come from the reference policy itself, we show that it covers the optimum up to a power law whose exponent grows as $H/β$; in the same temperature range, this power law alone gives the estimator a sample complexity polynomial in $1/ε$, with no coverage assumption. For $β\geq H$ the power can also increase the sample size needed by a factor polynomial in $1/ε$. Finally, the plug-in estimator, which is minimax optimal with a generative model, can fail on reference data at small temperatures when many actions are rarely seen, while the pessimistic estimator does not.

cs.LG

新论文 · v1arXiv:2611.17600

LightTAR: Light and Fast Target-Aware Representations for Multimodal Tabular Learning

Eilam Shapira, Alan Arazi, Shoham Grunblat, Roi Reichart

原文摘要

Multimodal tabular learning involves classification and regression on tables combining structured numerical and categorical features with unstructured text and images. These unstructured fields are commonly represented by pretrained encoder embeddings reduced with principal component analysis (PCA). This reduction prioritizes input variance but may overlook information relevant to predicting the target. We introduce LightTAR, a simple, inexpensive method that adapts dimensionality reduction to the prediction task. LightTAR augments the standard PCA representation with a few target-aware features extracted jointly from the final and an intermediate encoder layer using regularized partial least squares (PLS). We evaluate LightTAR with 21 tabular models, including tree ensembles, neural networks, and tabular foundation models. LightTAR improves average predictive performance for all models and, averaged across models, improves prediction on almost every MulTaBench dataset. These gains come with a small increase in total runtime relative to PCA representations. LightTAR integrates directly into existing encoding pipelines, improving tabular learners' predictive performance without modifying the encoder or the learner.

cs.LG · cs.CL · cs.CV

新论文 · v1arXiv:2611.17598

X-Nav Technical Report: Real-World Long-Horizon Indoor-to-Outdoor Navigation on a Humanoid Robot

Jiaxin Huang, Chengxing Lin, Xiaoyu Luo, Sirui Chen, Jingyu Guo, Aolin Li, Kairui Chen, Rui Zhen 等 13 位作者

原文摘要

Long-horizon navigation from indoors to outdoors remains an open problem for humanoid robots: one instruction covers rooms, doors and roads, and the camera moves with the gait. Worken, AGIBOT's measure of productive robot work, depends on robots that reach their workplaces autonomously. We present X-Nav, to our knowledge the first system to take a bipedal humanoid through a language-instructed long-horizon mission from indoors to outdoors. X-Nav covers a pipeline from data to deployment. Its data engine collects navigation data in open-source scenes, in digital twins of deployment sites and on real robots. Its navigation model, MELIOR, takes only RGB images from the robot's built-in camera and comes in two versions: MELIOR-A predicts actions, and MELIOR-N predicts a navigation goal. Its agentic workflow, X-Navarch, turns a brief human instruction into sub-tasks and deploys MELIOR on the robot with explicit target navigation and obstacle avoidance. On the Lingxi X2, X-Nav completes a mission of about a hundred metres from an office lobby to a store in a neighbouring building; LiDAR and odometry are used for target approach and low-level control. The mission is completed with both versions of MELIOR: the robot follows a change of destination and walks through a door from the lobby to the road, and with X-Navarch it also passes static and dynamic obstacles on the route. During deployment, MELIOR is adapted with data collected in a digital twin of the site; the adaptation raises its closed-loop success rate in the twin, and the adapted model is then deployed on the robot at the real site. In addition, MELIOR achieves competitive results on the R2R-CE and RxR-CE benchmarks.

cs.RO · cs.AI · cs.CV

新论文 · v1arXiv:2611.17596

Tractable Importance Sampling for Monte Carlo Privacy Accounting

Galen Andrew, Arun Ganesh

原文摘要

Evaluating tight Differential Privacy (DP) parameters for DP matrix factorization with privacy amplification by sampling requires analyzing high-dimensional, correlated Gaussian mixtures. Because integrating these mixture distributions to find the exact privacy failure probability ($δ$) is analytically intractable, these mechanisms are typically analyzed using Monte Carlo accounting. However, standard sampling suffers from prohibitive variance when estimating small $δ$ associated with strict privacy guarantees, which in turn requires a large sample complexity, making Monte Carlo accounting inefficient. We demonstrate that Monte Carlo accounting can be made much more efficient via importance sampling. We give a shifted importance sampling proposal distribution parameterized by a stretch factor $α> 1$. We prove a closed-form bound $B(α) < 1$ on the support of the importance-weighted estimator that holds identically for both directions of the hockey-stick divergence, show that $B$ is log-convex so that the optimal $α$ is a one-dimensional root-finding problem, and show that the resulting estimator seamlessly plugs into the KL-Hoeffding verification protocol used by past work, with a sample complexity reduced by exactly a factor of $1/B(α^*)$.

cs.CR · cs.LG

版本更新 · v2arXiv:2610.24349

FlashRT-PAI: Fast Kernels Are Not Fast Pipelines for Physical AI

Liang Su, David Chen, Jason Li, Kevin Guo, Teresa Lv

原文摘要

Vision-language-action (VLA), video-action (VA), and world action models (WAMs) form the backbone of modern physical AI systems, yet their real-time single-batch inference remains a defining bottleneck. Latency is dominated not by individual kernel computation, but by system-level overhead: thousands of short kernel launches, intermediate data movement across operator boundaries, and redundant recomputation across consecutive decisions. General-purpose graph compilers cannot address these costs, as they optimize individual calls in isolation and lack visibility into cross-call state semantics and global pipeline structure. We present FLASHRT-PAI, a low-latency inference engine built around model-hardware pipeline co-design rather than isolated kernel optimization. FLASHRT-PAI combines three core mechanisms: deep kernel fusion with launch overlap eliminates inter-operator boundaries and launch gaps; an execution contract formalizes the pipeline's stage DAG, update rates, and named buffers to enable cross-decision state reuse and selective refresh; and execution-state capsules enable fast snapshot, restore, and fork of persistent pipeline state. The system also supports drop-in acceleration of unmodified PyTorch models via parity- and latency-gated kernel replacement. Evaluation covers six policies on RTX 5090 and four on Jetson AGX Thor. At matched 16-bit precision, FLASHRT-PAI is 1.3-3.7x faster than official compiled paths. Combining pipeline co-design with reduced precision delivers 3.38-12.66x speedup over official eager implementations. In LIBERO-10, four-bit pi0.5 achieves 94.04% success on Thor and 93.32% on RTX 5090, compared with 92.4% for the measured official FP16 reference. FLASHRT-PAI is open source, distributed via Hugging Face Kernel Hub, and its opt-in pi0.5 kernels have been adopted into the LeRobot framework.

cs.AI

版本更新 · v3arXiv:2609.07729

Attributing Cohen's d: Training Data Attribution for Disease-Related Effects in Normative Age Biomarkers

Jakob Snel, Marc-Andre Schulz

原文摘要

Normative age models are trained to predict chronological age in a nominally healthy cohort. In patients, the gap between predicted and chronological age is read as disease risk. We attribute the disease-related effect size of the age gap directly to individual training samples, rather than using a prediction-level loss as the attribution target. For the Cohen's d effect size, the resulting closed-form influence functional, validated against leave-one-out retraining, ranks training samples by their effect on held-out case-control separation. Across four diseases and two biomarker modalities in UK Biobank, removing the 10% most influential training samples raises held-out disease-related effect size in every seed. It more than doubles the metabolomic-age effect for type 2 diabetes and raises the brain-age effect for multiple sclerosis by roughly a third. Random removal leaves effect size flat even at 50% removal, confirming the gain comes from which samples are removed, not how many. Flagged subjects, missed by diagnosis-based exclusion, show elevated cardiometabolic markers the model never sees, consistent with subclinical contamination of the reference cohort. For type 2 diabetes, where the method gains most, the marker recovered is HbA1c, the standard measure of blood sugar control. We release our influence-function package pyinfluence.

cs.LG · q-bio.QM

版本更新 · v2arXiv:2611.00406

TAP-Out: Tracking Any Point in 360° via Outpainting

Finlay G. C. Hudson, James A. D. Gardner, William A. P. Smith

原文摘要

Tracking points that leave the camera's field of view (FoV) requires reasoning about unobserved scene dynamics, a fundamentally different problem from tracking visible correspondences. We present TAP-Out, an offline method that makes this out-of-view (OOV) reasoning explicit by leveraging generative video outpainting as a world-model prior. Given a narrow-FoV input video, TAP-Out outpaints a complete 360° equirectangular video, locates target objects within the panoramic scene via video object segmentation, and synthesises object-centred perspective crops on which standard in-frame trackers operate. This decoupled design composes off-the-shelf foundation models without requiring task-specific architectural modifications. On the TAPVid360 benchmark, TAP-Out achieves state-of-the-art performance, improving OOV tracking accuracy by approximately 3x over the best prior method. We demonstrate its flexibility on out-of-frame object segmentation and detection and we provide detailed diagnostic analyses across motion regimes, static versus dynamic content, and hallucination quality, characterising when and why the approach succeeds or fails. Code is publicly available at https://github.com/finlay-hudson/TAP-Out

cs.CV

新论文 · v1arXiv:2611.17589

Small Open-Weight Language Models for Crowd Decision-Making: Capabilities and Limits of Output-Level Fusion and Linguistic Aggregation

David Herrera-Poyatos, Cristina Zuheros, Carlos Peláez-González, Andrés Herrera-Poyatos, Rosana Montes, Francisco Herrera

原文摘要

Small open-weight language models can run crowd decision-making (CDM) on an institution's own infrastructure, which keeps the checkpoint, the inference settings and the opinions under its control. Their judgements vary across repeated queries; whether repeating and aggregating them, often through the 2-tuple linguistic representation, settles the decision rather than the labels is untested. Five open-weight models of 8-36B total parameters on one consumer GPU and a hosted reference classify 474 TripAdvisor reviews of four London restaurants ten times each; three open-weight models repeat the protocol on 600 reviews of twelve Madrid restaurants. Continuous evaluations preserve resolution: snapping each review leaves 5 distinct values against 21-65, and snapping at every level at most two collective values. Sensitivity to the participants persists once run-to-run variability is reduced: in London, redrawing the reviewers keeps the winner on 47-81% of replicas against 86-100% for redrawing the runs. With simultaneous intervals over every pair of alternatives under crowd resampling, no system on either corpus names a single winner: in London every system keeps all four restaurants as candidates, and in Madrid the candidate sets hold five to eight of the twelve. Extra runs and output-level fusion stabilize the winner but do not remove that sensitivity, and the fusion does not beat its best member. The three larger open-weight models name the hosted reference's London winner and at least match its per-review correlation with the reviewers' stars, so small open-weight models can support local CDM, reported as candidate sets with margins rather than forced rankings.

cs.CL

新论文 · v1arXiv:2611.17586

Where to Predict in Retinal OCT? Anatomy-Guided Target Selection for I-JEPA

Yunchu Feng, Wai Tak Lau, Kaveri A. Thakoor

原文摘要

Self-supervised learning aims to learn meaningful representations from unlabeled data, potentially allowing models to capture subtle anatomical variations across patients and reduce the amount of labeled data needed. Image-based Joint-Embedding Predictive Architecture (I-JEPA) learns by predicting target-region representations from visible context rather than reconstructing pixels. However, its uniform target placement does not account for the anatomical structure of retinal OCT, where retinal tissue occupies only part of each B-scan. We investigate anatomy-guided target selection using image intensity and predicted retinal segmentation while retaining the I-JEPA architecture and latent prediction objective. We compare anatomy-guided rectangular targets, anatomy-shaped targets and coverage-constrained placement with uniform masking, and evaluate the learned representations using linear probing for glaucoma classification on FairVision. We found that rectangular targets guided by image intensity achieved the highest linear-probe AUC in our comparisons. We show that anatomically meaningful target selection is promising for retinal OCT representation learning, while greater anatomical specificity did not consistently improve performance.

cs.CV

新论文 · v1arXiv:2611.17585

Rethinking Visual Reasoning in Text-to-Image Reward Modeling

Shiye Su, Xiaohan Wang, Ludwig Schmidt, Serena Yeung-Levy

原文摘要

Reward models are a core component of modern text-to-image systems, serving as the objective for inference-time selection and RL-based post-training. A key prerequisite is visual understanding: to judge prompt adherence and ground aesthetic preferences, a reward model must bind attributes, resolve relations, and count reliably. We find that widely-used reward models perform poorly on targeted visual reasoning evaluations. Even when starting from strong vision-language backbones, preference fine-tuning erodes these capabilities, and continual-learning baselines fail to mitigate a steep tradeoff between visual reasoning and preference accuracy. We propose a data-centric intervention that injects visual reasoning supervision into preference learning by flipping the pairing axis. To complement conventional human preference data (paired images under a shared prompt), we convert existing visual reasoning datasets into preference pairs that share the same image but differ in prompt, and train on the mixture. This yields a Pareto improvement in both visual understanding and preference prediction. Across inference-time best-of-n selection and RL post-training, the resulting reward model improves text-to-image generation on challenging compositional prompts (GenEval) by an average of 4.0% and 5.5% respectively, outperforming HPSv3 which was trained on 6x more preference pairs.

cs.CV · cs.AI

版本更新 · v2arXiv:2608.22228

Prompt-Based Abstention Fails Under Misleading Context: A Controlled Study of Small Frozen RAG Models

Yohanes Andre Setiawan

原文摘要

Missing and misleading evidence are not the same problem in retrieval-augmented generation (RAG), but prompt-based abstention treats them alike. Small frozen models abstain when context is absent, not when it is misleading. We introduce GRAB-RAG (Graded Abstention Benchmark for Retrieval-Augmented Generation), a paired benchmark that tests the same questions across four context conditions (supportive, degraded, missing, and misleading) in Natural Questions and HotpotQA. In the misleading condition, we edit a gold passage to support a wrong answer and place it among other retrieved passages. We test five abstention policies on three small frozen models (3.8B-8B) across two English QA benchmarks. Models abstain reliably when evidence is missing, but under explicit abstention prompting still answer 41.6% of misleading questions, with 63% of those answers echoing the planted wrong entity. Chain-of-thought provides little additional benefit. A generator-side conflict check cuts the rate to 13.3% but discards many correct answers, while a natural language inference (NLI) verifier recovers that coverage but fails when parametric memory and the misleading passage agree on the same wrong answer. Prompt-based abstention asks whether context is sufficient, not whether it is correct. The two-pass checks are diagnostic baselines, not deployable mitigations.

cs.IR

新论文 · v1arXiv:2611.17583

When Does Less Information Pay? An Exact Finite Atlas of Representation, Payoff, and Cost

Aleksandr M. Shtraukh-Adamsky, Vladimir M. Adamsky, Irina G. Voloshina, Artemiy D. Spitsyn

原文摘要

A representation that retains one bit about a four-state world can solve a task perfectly, while another that retains 1.5 bits on average reaches only 75% accuracy. The difference lies in which distinctions each representation preserves. We examine this relationship in a uniform four-state world by enumerating all admissible representations and binary action rules. Constructions are compared both by replacing one component while holding the others fixed and by choosing the best action rule separately for each module. We consider three tasks, their mixtures, and four separate hypothetical cost tariffs. In some comparisons, the representation retaining less world information has the higher payoff; in others, the more informative representation does. There are also pairs with different information amounts but equal payoff--that is, equal task accuracy. If a representation preserves all existing distinctions and adds new ones, its best zero-cost payoff cannot decrease. Lower-information advantages therefore depend on which distinctions are retained. Equal counts of opposing advantages under representation replacement follow from coding symmetry. In broader comparisons, the shares depend on how repeated realizations are counted. A different task set or resource price can change which member of a fixed pair performs better. These results are exact, complete censuses of the specified finite sets of constructions and comparisons. The atlas provides a reproducible worked model for examining how action constraints, counting rules, and cost assumptions affect comparisons. It illustrates established distinctions between information quantity and decision value rather than proposing a new general law of their relation. $I(W;R)$ measures retained world information, and payoff measures task accuracy.

cs.AI

版本更新 · v2arXiv:2611.00430

Real-VAS: a Real-World Video Amodal Segmentation Dataset

Finlay G. C. Hudson, William A. P. Smith

原文摘要

Amodal video object segmentation is fundamentally limited by the absence of datasets that combine real-world complexity with precise ground-truth annotations. To address this, we present Real Video Amodal Segmentation (Real-VAS), a new large-scale, zero-shot evaluation dataset built from composited real-image video. We introduce a novel data generation pipeline that enables the creation of programmatically consistent amodal ground truth for real-world footage, in which the object mask, occlusion boundary, and composited image are all derived from a single unoccluded source, without relying on human estimation or expensive 3D reconstruction. Our dataset is structured into two challenging scenarios: dynamic Occlusion, created by compositing two clips of moving objects, and a unique Container category featuring complex, physically constrained interactions. These container scenarios--On Surface Containment, Articulated Containment, and Mobile Containment--allow us to generate precise ground truth by simulating an object's motion based on its container's tracked transformation. As a result, Real-VAS provides a diverse and challenging benchmark for evaluating amodal segmentation models on realistic video with the precision of synthetic data. The dataset and generation code are publicly available at https://finlay-hudson.github.io/real_vas

cs.CV

版本更新 · v2arXiv:2608.18389

Programmable Distributions of Semantics-Preserving Variants for Evaluating Code Agents

Hasan Najib Mahmud, Shreya Gupta, Isha Chaudhary, Nathaniel Enis, Ravi Mangal, Gagandeep Singh, Corina Pasareanu

原文摘要

Code agents are increasingly deployed to resolve real software issues, yet their reliability under superficial code variations remains poorly understood. We evaluate whether coding agents that repair repository-level issues remain reliable when the surrounding codebase is rewritten into a semantically equivalent form. We sample semantically equivalent variants of a repository from an explicitly defined distribution and compare the agent's resolve rate on them with that on the original. Different distributions reveal different aspects of an agent, such as its sensitivity to routine refactoring or to misleading code near the location of a fix. We develop VDL, a probabilistic programming language for expressing such distributions, whose programs randomly apply semantics-preserving transformations (SPTs) to a repository. For our evaluation, we use it to define a generic distribution that is not targeted at any particular failure mode. We evaluate two agentic scaffolds (mini-SWE agent and OpenCode), each backed by one of four frontier models (Claude Opus 4.5, Kimi K2.5, MiniMax M2.5, and Qwen 3.6-27B), on instances from SWE-bench Verified and SWE-bench Pro. Resolve rates drop by up to 6.7 percentage points, significantly in 6 of 16 configurations. No ranking of models by robustness holds across scaffolds, revealing a jagged robustness frontier.

cs.AI · cs.SE

版本更新 · v3arXiv:2510.18691

Investigating LLM Capabilities on Long Context Comprehension for Clinical Question Answering

Feras AlMannaa, Talia Tseriotou, Jenny Chim, Maria Liakata

原文摘要

Despite the impressive reported performance of Large Language Models (LLMs) on medical Question-Answering (QA), evaluation still overrelies on multiple-choice and extractive settings primarily assessing textbook knowledge, with limited work on open-ended QA over patient records. We present a comprehensive study investigating LLM long-context (LC) comprehension on longitudinal, patient-centric Electronic Health Records (EHRs) across multiple datasets, models and metrics, covering extractive, multiple-choice, and open-ended generative QA under different context formulation and Retrieval Augmented Generation (RAG) settings. We show that medically fine-tuned models offer no measurable benefit over base counterparts, smaller models exhibit stronger memorization, and longitudinal supplementary notes aid generative QA. Reasoning models gain on complex multi-note QA under LLM Correctness but not stricter reference-aligned metrics, while hybrid RAG with reranking outperforms Full-Context (FC) prompting. A question-category analysis and multi-faceted metric evaluation further reveal where RAG and FC excel and struggle, with RAG falling behind in longitudinal reasoning, thus uncovering systematic failures pointing to the need for reasoning-appropriate metrics.

cs.CL

新论文 · v1arXiv:2611.17577

Fair and Efficient Distributed EV Charging Coordination Using Threshold Cryptography and Machine Learning

Md. Mainul Islam, Muhammad Ismail, Hasan Kurban, Abdulrahman Takiddin, Erchin Serpedin

原文摘要

Distributed coordination of electric vehicle (EV) charging with the alternating direction method of multipliers (ADMM) keeps EV constraints private but assumes that every participant follows the algorithm. This assumption fails in practice, where a biased operator or a man-in-the-middle (MITM) attacker can send inconsistent signals to different EVs, and compromised EVs can deviate from their prescribed updates. We first prove that attacks by such arbitrarily malicious, i.e., Byzantine, participants do not prevent convergence and therefore remain undetected, and derive the most damaging attack in closed form. We then propose a framework that combines a threshold-signed broadcast with upload recomputation, which detects every deviation without false alarms at linear instead of quadratic communication cost. To keep the optimization running, each affected EV recovers locally with a certified, fleet-aware machine learning price forecast, and we prove that the optimization remains convergent with a recovery cost bounded by the shape of the forecast error. Simulation results with real-world load, price, and charging data show that a biased operator raises the annual cost of disadvantaged EVs by 132.7% and lowers that of advantaged EVs by 49.3% without changing the aggregate load, whereas the defense limits all cost changes to 2.6%. An external MITM attacker raises the cost of its victims by 57.6%, the peak load by 5.8%, and the grid cost by 15.8%, whereas the defense keeps them within 1.1%, 0.2%, and 0.3% of benign operation. With up to 70% of the EVs attacked, the recovery converges in at least 97.8% of the runs.

cs.LG · cs.CR · eess.SY

版本更新 · v2arXiv:2606.30237

Comparing Human and Automatic Recognition of Dutch Dysarthric Continuous Speech: A Case Study

Yuanyuan Zhang, Dimme de Groot, Jorge Martinez, Odette Scharenborg

原文摘要

In our goal to develop personalised dysarthric speech recognition (DSR) models, this study compared the recognition performances of human listeners and those of three state-of-the-art, off-the-shelf ASR systems (Whisper-large-V3, Google Chirp 3, and Omnilingual) on the recognition of Dutch continuous read and spontaneous speech from a single speaker with severe dysarthria. Results showed that both humans listeners and the three off-the-shelf ASR systems exhibit word error rates (WER) exceeding 70% on average, indicating that DSR is highly challenging for both humans and ASR systems. Fine-tuning on the dysarthric speech significantly reduced WER. Although overall WERs are still quite high (>23%), the personalised DSR models outperformed the human listeners, and performance is getting closer to being useful for supporting day-to-day communication of dysarthric speakers. Future research should focus on improving personalized DSR on spontaneous speech and longer utterances in the case of read speech, with a specific focus on particular phonemes.

cs.CL

新论文 · v1arXiv:2611.17576

Beyond a Single Sign: BiTE for Branch-Preserving Binary Threshold Expansion

Siddharth Gupta, Mikail Yayla, Akash Kumar

原文摘要

Binary neural networks (BNNs) reduce model storage and replace many arithmetic operations with efficient bitwise computations, but the single-sign activation limits their representational capacity. Each preactivation direction is compared with one threshold and reduced to one of two values, discarding variation within each sign region. We propose \textbf{BiTE} (Branch-Preserving Binary Threshold Expansion), which evaluates each feature direction against $K$ ordered thresholds and retains the resulting binary responses as separate inputs to the following learned operator. BiTE can distinguish up to $K+1$ ordered activation states while allowing the following operator to learn a separate transformation for each threshold response. It also enables a controllable trade-off between the number of independently constructed feature directions and the number of ordered responses extracted from each direction. On BinaryViT, BiTE improves mean final-epoch Top-1 accuracy from $74.60\pm0.30\%$ to $75.83\pm0.15\%$ on CIFAR-10 and from $42.80\pm0.46\%$ to $44.07\pm0.32\%$ on CIFAR-100, with nearly unchanged trainable parameter counts. A matched uniform-collapse control reaches $75.33\pm0.15\%$ on CIFAR-10, supporting the value of retaining the threshold responses separately relative to the evaluated scalar collapse. Binary VGG-16 experiments further show a favorable observed parameter--accuracy trade-off, including a configuration that achieves $0.64$ percentage points higher mean accuracy with $21.8\%$ fewer parameters than conventional width scaling. Overall, these results support branch-preserving threshold expansion as an effective way to improve binary representations and the observed parameter--accuracy trade-off.

cs.LG

新论文 · v1arXiv:2611.17575

Natural Swing-Up of an Underpowered Double Pendulum

Arne Sachtler, Manuel Keppler, Florian Loeffl, Alin Albu-Schäffer

原文摘要

Robots with weak motors must ride their natural dynamics rather than fight them. We swing up a direct-drive planar two-link robot, whose motors deliver only about 6\,\% of its shoulder gravity torque, along its nonlinear normal modes, two-dimensional invariant manifolds consisting of periodic orbits. On a non-strict mode, injecting energy necessarily moves the state off the manifold, so pumping and stabilization compete for the same bounded torque. We split the input torque into a power-neutral steering component and a scalar pumping input, the only component that changes the energy. The allocation then becomes a linear program in two variables: maximize pumping subject to the torque limits and one control-Lyapunov constraint. The constraint uses the model-based rate of the error, so no stabilizing feedback gain is needed. Experiments in hardware show swing-ups along both modes and compare the allocation with summing-and-clipping and with prioritized stabilization.

cs.RO · eess.SY

版本更新 · v2arXiv:2611.03940

Coordinate Frame Recovery, Per Class NMS and Null Return Modelling for Sim2Real V2X LiDAR Detection

Aman Upganlawar

原文摘要

We describe our approach to the V2X-Real track of the UrbanTwin Sim2Real LiDAR Challenge, in which a 3D object detector must be trained exclusively on synthetic data and evaluated on unseen real roadside LiDAR from an Ouster OS1-64 unit, while the synthetic point clouds are simultaneously scored for distributional realism. Our final entry scores 0.7457 combined, from a detection mAP of 0.4382 and a realism score of 0.9251. Three findings account for most of our result, and all three were diagnostic rather than architectural. First, a coordinate frame mismatch between our predictions and the evaluation ground truth was suppressing detection AP entirely; recovering it took our score from 0.3520 to 0.5147 in a single submission. Second, a single global non maximum suppression threshold was destroying genuine pedestrian detections, because pedestrians occur in groups and a 0.3 IoU threshold applied to a 0.7 m box suppresses distinct neighbouring people; correcting this per class raised overall mAP from 0.0751 to 0.4295. Third, approximately 31 percent of the real reference point cloud consists of null returns encoded as the literal point (0, 0, 0), a format feature entirely absent from our synthetic clouds, which pinned Earth Mover's Distance near its floor until we modelled it. Negative results are reported at equal length, and Section 9 gives a full account of data provenance and compliance. That section also accounts for every mechanism contributing to the submitted score.

cs.CV · cs.RO

新论文 · v1arXiv:2611.17573

CamSpokenNLI: A Pilot Resource for Cross-Variety Transfer Evaluation in Cameroon Pidgin English and Camfranglais

Tiani Pekins Ebika

原文摘要

Cameroon Pidgin English (a West African Pidgin English variety) and Camfranglais are widely used in Cameroon, yet to our knowledge, neither has a dedicated spoken Natural Language Inference (NLI) resource. We introduce CamSpokenNLI, a pilot dataset of 114 scenario-matched spoken NLI items in each variety (228 in total), recorded by one native speaker from Buea. Each item pairs premise audio and text with a hypothesis, an NLI label, and English glosses. The author assigned labels, so the dataset is an exploratory resource rather than a mature benchmark. The NLI experiments use the reference text; the audio is evaluated separately. XLM-RoBERTa baselines with five-fold item-level cross-validation reach 44.7% on Pidgin and 46.2% on Camfranglais (chance 33.3%). Still, neither is statistically distinguishable from hypothesis-only input, so hypothesis wording may account for much of the accuracy. Under disjoint-item cross-variety transfer, where a test item's counterpart is never seen in training, accuracy is 40.1% (Pidgin to Camfranglais) and 45.6% (Camfranglais to Pidgin). A counterpart-leak audit raises this to 60-64%, illustrating how counterpart leakage can overstate transfer. Zero-shot Whisper transcription of the recordings performs poorly: the best word error rates are 66.6% for Pidgin and 45.5% for Camfranglais, and with automatic language identification, Whisper Large-v3 labels 93 of 114 Pidgin recordings as Yoruba or Swahili. Together, the results suggest that such resources need disjoint-item evaluation and hypothesis-only controls, and that speech recognition must be adapted to these varieties before spoken NLI can be tested end to end.

cs.CL · cs.SD

版本更新 · v3arXiv:2512.20777

Improving Matrix Exponential for Generative AI Flows: A Taylor-Based Approach Beyond Paterson--Stockmeyer

Jorge Sastre, Daniel Faronbi, José Miguel Alonso, Peter Traver, Javier Ibáñez, Nuria Lloret

原文摘要

The matrix exponential is a fundamental operator in scientific computing and system simulation, with applications ranging from control theory and quantum mechanics to modern generative machine learning. While Padé approximants combined with scaling and squaring have long served as the standard, recent Taylor-based methods, which utilize polynomial evaluation schemes that surpass the classical Paterson--Stockmeyer technique, offer superior accuracy and reduced computational complexity. This paper presents an optimized Taylor-based algorithm for the matrix exponential, specifically designed for the high-throughput requirements of generative AI flows. We provide a rigorous error analysis and develop a dynamic selection strategy for the Taylor order and scaling factor to minimize computational effort under a prescribed error tolerance. Extensive numerical experiments demonstrate that our approach provides significant acceleration and maintains high numerical stability compared to existing state-of-the-art implementations. These results establish the proposed method as a highly efficient tool for large-scale generative modeling.

cs.LG · math.NA

版本更新 · v2arXiv:2611.03936

Geometric Refinement and Distributional Alignment for Sim2Real Roadside LiDAR Detection

Aman Upganlawar

原文摘要

We describe our approach to the LUMPI track of the UrbanTwin Sim2Real LiDAR Challenge, in which a 3D object detector must be trained exclusively on synthetic data and evaluated on unseen real roadside LiDAR, while the synthetic point clouds are simultaneously scored for distributional realism. Our final entry scores 0.4793 combined, from a detection mAP of 0.1333 and a realism score of 0.9126. Our contribution is not a new architecture. Working from a PointPillars baseline, we obtained most of our gains from three sources. First, a diagnosis that small object failure in roadside LiDAR is a localisation problem rather than a detection problem, which led to a parameter free geometric refinement worth 86 percent relative on pedestrian AP. Second, an audit of our own post processing which revealed that a shipped rule was destroying valid predictions, the repair of which produced our single largest detection gain at 335 percent relative on the Van class. Third, treating the realism metrics analytically rather than as a black box, which showed that one of the four components is a closed form function of nine summary statistics. We also report an experimental methodology that we believe is the more transferable result. Because mean AP decomposes into independent per class terms, a single submission can carry several mutually controlled experiments, and a submission assembled entirely from previously measured components has a score that is known before it is sent. We predicted our final score to four decimal places prior to submission. We report negative results at equal length, and give a full account of data provenance and compliance in Section 8, including a full account of every mechanism contributing to the submitted score.

cs.CV · cs.RO