跳到正文
研踪PaperTrail

YOUR RESEARCH, IN SIGHT

让下一篇重要论文，不再错过。

覆盖计算机全学科，以 AI 为快捷入口。关注方向、作者和关键词，把发现变成持续的研究积累。

部分采集异常
正在读取账号与研究记录…

全站更新概况

+1072新增论文
+845其中 AI 相关
+94已有论文修订
+100新增 AI 导读

arXiv 最近完成扫描：2026-11-09 21:43 · 全站最近入库：2026-11-09 13:22各信源进度 →

统计口径与数据新鲜度

统计 2026-11-08 22:45 — 2026-11-09 22:45（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：2026-11-09 21:43最近发现新增版本：2026-11-09 13:22库内论文最新更新时间：2026-11-07 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。

18,172 已收录论文15,506 AI 相关4,527 有修订112 HF 社区发现

浏览论文库

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 18,172 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-09 21:40（北京时间）
新论文 · v1arXiv:2610.35566

Copying As Defense: Protecting Images from Editing By Eliciting Model Copying

Salma Abdel Magid, Sanghyuk Chun, Eric Slyman, Kushal Kafle, Olga Russakovsky

原文摘要

Image editors built on diffusion transformers (DiTs), such as Flux Kontext, make it possible to turn a single photograph into a convincing manipulated image, including harmful and non-consensual edits. Proactive defenses aim to prevent such manipulations by adding an imperceptible perturbation before an image is shared. Existing defenses for DiT-based in-context editors make the output attend less to the protected reference image. But this leaves the model free to follow the edit instruction using the text and its own generated tokens, often producing a new, weakly grounded image rather than preventing the edit. We find that the opposite strategy is more effective: instead of leading the model to ignore the reference, we encourage it to copy it. Through controlled attention interventions, we show that concentrating target-token attention on the reference image in only a few early blocks is sufficient to collapse the editing behavior, causing the model to reproduce its input, often as a lossy copy, rather than execute the requested transformation. Based on this insight, we introduce Copying-as-Defense (CaD), which maximizes per-token target-to-reference attention mass in early transformer blocks. Since attention rows are normalized, increasing reference attention suppresses the attention allocated to the text and generated canvas. Restricting the objective to early blocks also lets us truncate the backpropagation path, keeping optimization cheap. Across editors and benchmarks, CaD reduces average edit success from 81.4% on unprotected images to 24.0%, a 70.6% relative reduction, and by a further 44.3% relative to DeContext, the strongest baseline (43.0%). CaD is also roughly 1.8-2.0x faster per optimization step than comparable methods.

cs.CV

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35564

A Unified Queueing-Based Framework for Comparative Evaluation of Serverless Autoscaling Strategies under Bursty Workloads

Zahra Zafarzade, Ehsan Ataie

原文摘要

Serverless computing platforms must preserve quality of service (QoS) while adapting to highly bursty and unpredictable workloads. To cope with workload fluctuations while satisfying QoS requirements and service-level agreements (SLAs), these platforms rely on automatic resource scaling. Commercial function-as-a-service (FaaS) platforms employ different autoscaling strategies to dynamically adjust the number of active function instances and the available processing capacity. In this work, three representative autoscaling strategies are modeled and compared: scale-per-request (SPR), concurrencylevel scaling (CLS), and metric-based scaling (MBS). To enable systematic evaluation independent of proprietary platform-specific implementations, we develop a unified modeling framework that combines analytical queueing network (QN) modeling with queueing Petri nets (QPN) implemented in Java Modeling Tools (JMT) to evaluate autoscaling behavior under two-phase hyper-exponential (H2) arrivals. Evaluation is conducted using four system-level metrics: throughput, mean response time, system population, and system power. The framework captures dynamic function instance (FI) activation, concurrency constraints, and queue buildup under saturation. Comparative analysis reveals a clear trade-off between peak performance and stability: SPR provides the lowest delay under moderate load but becomes fragile near capacity limits; CLS is governed by concurrency tokens and is highly sensitive to configuration under bursty conditions; whereas MBS provides the most stable QoS across the workload range by keeping utilization below a critical region and preventing explosive queue growth under critical load.

cs.DC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35563

High-resolution tactile observation makes hidden racket-control mechanisms learnable

Haoyang Tian, Zhengwei Li, Houcheng Li, Xu Yang, Peigeng Gu, Xinyue Wang, Tairan Zhu, Shuo Jiang 等 9 位作者

原文摘要

Experts can master fine hand-object manipulation skills, yet limited conscious perception of rapid tactile interactions hinders their ability to articulate the underlying control mechanisms. Badminton racket control requires rapid, fine hand-force coordination, providing a demanding testbed for uncovering these mechanisms. An 864-node tactile-inertial glove with synchronized 100-Hz acquisition recorded over 14,000 strokes from more than 70 participants across five skills. Inspired by sensorimotor control, our mechanics-grounded two-layer framework inferred object-driving and grip-stabilizing control from tactile history, while resolving their spatiotemporal force realization. We found that higher proficiency was associated with selective rather than extensive control: greater skill specificity, anticipatory regulation, focal distal modulation and lower geometric torque cancellation, while spatiotemporal coordination showed a non-monotonic acquire-refine pattern. We translated these mechanisms into personalized guidance targeting force control and force-motion coordination. In a randomized cohort, mechanism-guided learners gained 0.72 expert-rated skill levels versus 0.05 with level-only feedback, making control mechanisms actionable.

cs.LG · cs.HC · cs.RO · eess.SP

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35562

Beyond Being Accurate: Implications for Using Natural Language Descriptions for Accessible 3D Modeling with~Blind and Low Vision Makers

Gene S-H Kim, Nicholas Oskiper, Sean Follmer, Megan Hofmann

原文摘要

People who are blind or have low vision (BLV) are largely excluded from 3D modeling tools that rely on visual representations of geometry. Natural language descriptions offer a potential path to accessibility, but it is unclear whether language alone can support the deep spatial understanding that editing, not just browsing, requires. Advances in generative AI offer the promise of automatic model description, yet descriptions suited to BLV modelers are structurally distinct from alt-text and from existing 3D model captioning datasets, and the domain-specific requirements for generating them are not well understood. We address this through two coordinated studies: an autoethnographic pilot in which the blind first author worked with 12 sighted participants to develop a taxonomy of descriptive strategies, and a Wizard-of-Oz main study that applied those strategies as idealized scripts with 10 BLV participants. We identify a systematic use-to-editing gap: descriptions that support accurate but incomplete mental models are insufficient for specifying modifications, because editing requires knowing the state of features the user did not know to ask about. We show that this gap produces four failure types (accuracy, completeness, expression, and navigation) that accuracy-focused evaluation frameworks cannot detect. We propose a two-phase evaluation framework that diagnoses the failure type a description system exhibits and guides developers toward targeted design interventions, thereby providing a methodological standard for assessing whether description-based systems genuinely support BLV modelers as creators.

cs.HC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35558

EXCISE: Locating and Excising Distracting Sentences in Medical Question Answering

Yerim Hong, Sohyun Son, Heeseung Chung, Mujeen Sung

原文摘要

Language models are increasingly considered for clinical decision support, where their answers should be grounded in clinical evidence and remain robust to irrelevant information. Prior work shows that a single irrelevant sentence can lower the accuracy of medical models, but not where inside the model its effect can be removed, or how to do so without knowing which sentence it is. We find that blocking attention to a known distractor restores the answer, whereas head masking can suppress correct answers. We also show that the model's hidden states mark the distractor as the sentence that least fits the case even when the model follows it. To remove a distractor's effect without knowing which sentence it is, we propose EXCISE, an inference-time method that blocks the sentence most influencing the answer, but only if the model's hidden states mark it as inconsistent with the case. On MedQA-TD (MedQA with Targeted Distractors), a benchmark we construct in which each distractor targets a specific wrong option, EXCISE gains 16.7-22.8 accuracy points at a 1.5-2.1-point cost on the same questions without distractors, outperforming prompting, rewriting, and steering baselines for all three models. Its gains extend to MedDistractQA and BiasMedQA, as well as chain-of-thought generation.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35556

LoopFold: Parallelizing Recurrent Depth in Looped Transformers

Feichen Song, Boyi Zeng, Yiqin Hao, Shixiang Song, He Li, Yizhou Yang, Zitong Wang, Yifan Liu 等 10 位作者

原文摘要

Looped Transformers increase computation per token by repeatedly applying a shared backbone, offering a way to improve language modeling and downstream task performance at a fixed backbone parameter count. However, each loop depends on the preceding loop, so increasing recurrent depth also increases decoding latency. Existing approaches include Parallel Loop Transformers, which reduce latency by overlapping loop computation across token positions, but yield rapidly diminishing marginal accuracy gains at larger depths. To address this challenge, we introduce LoopFold, which folds computation at multiple recurrent depths into parallel batches for the same token prefix. A lightweight predictor estimates inputs to later loops so they can start before earlier loops finish. A corrector then refines the outputs of these later loops using newly available results from earlier loops. All components are trained jointly using only the final next-token prediction loss. In experiments with 1.4B models, deeper LoopFold configurations achieve progressively higher average zero-shot accuracy across eight tasks, rising from 53.41% to 56.38%. Vanilla looped models and PLT show limited gains over the same depth range. Compared with a nine-loop vanilla model, LoopFold uses five sequential stages and additional parallel computation to reduce decoding latency by 46%.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v5arXiv:2608.00794

Measurement Without Validity: The Compounding Reliability Problem in Agentic AI Evaluation

William Caban

原文摘要

Agentic AI evaluation pipelines produce benchmark scores that justify deployment decisions, safety certifications, and regulatory compliance claims. No formal framework has characterized how validity degrades across these stages. We present a knowledge-based three-layer compounding validity model, $V_{\rm total} \leq V_1 \times V_2 \times V_3$, that captures multiplicative degradation across task generation ($V_1$), human-simulator calibration ($V_2$), and automated judgment ($V_3$). Applying it to five widely-used agentic benchmark configurations, we find $V_1$ and $V_2$ independently measurable and both degraded, but $V_3$ not currently quantifiable for any: three of the four benchmarks score with a single deterministic rule, so no second rater exists and reliability is undefined, and the fourth's partial LLM grader has no published grader-vs-human measurement. A structured survey of 55 published agentic evaluation papers finds that 26 of them report no chance-corrected inter-rater statistic of any kind, a determination four independent raters recover at Krippendorff's $α= 0.834$. On the broader judgment of whether a reported metric fits its rating design, approximately 80% of this purposive sample use structurally mismatched, incomplete, or absent inter-rater reliability (IRR) metrics (four-rater range 60.7%-84.4%). Both are consistent with systematic $V_3$ collapse. We further document $V_1$ failures (task validity flaws in 7 of 10 benchmarks) and $V_2$ miscalibration (up to 9 percentage points inter-simulator variance, with systematic demographic disparities for non-Standard American English speakers). We derive eight prescriptions grounded in psychometric science and domain-stratified reliability thresholds (intraclass correlation $\geq 0.70$; $α\geq 0.67/0.70/0.80$ by consequence level) that practitioners can apply immediately.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2601.01638

Interaction Improvement

Adrienne Lancelot, Giulio Manzonetto, Guy McCusker, Gabriele Vanoni

原文摘要

The relational semantics of linear logic is a powerful framework for defining resource-aware models of the $λ$-calculus. However, its quantitative aspects are not reflected in the preorders and equational theories induced by these models. Indeed, they can be characterized in terms of (in)equalities between Böhm trees up to extensionality, which are qualitative in nature. We employ the recently introduced checkers calculus to define a quantitative contextual preorder on $λ$-terms, and demonstrate that it coincides with the preorder associated to the relational semantics.

cs.LO · cs.PL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2601.06961

The Impact of Anisotropic Covariance Structure on the Training Dynamics and Generalization Error of Linear Networks

Taishi Watanabe, Ryo Karakida, Jun-nosuke Teramae

原文摘要

The success of deep neural networks largely depends on the statistical structure of the training data. While learning dynamics and generalization on isotropic data are well-established, the impact of pronounced anisotropy on these crucial aspects is not yet fully understood. We examine the impact of data anisotropy, represented by a spiked covariance structure, a canonical yet tractable model, on the learning dynamics and generalization error of a two-layer linear network in a linear regression setting. Our analysis reveals that the learning dynamics proceed in two distinct phases, governed initially by the input-output correlation and subsequently by other principal directions of the data structure. Furthermore, we derive an analytical expression for the generalization error, quantifying how the alignment of the spike structure of the data with the learning task improves performance. Our findings offer deep theoretical insights into how data anisotropy shapes the learning trajectory and final performance, providing a foundation for understanding complex interactions in more advanced network architectures.

stat.ML · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35551

Balanced Sampling in Stochastic Multi-objective Bandits

Mengfan Xu

原文摘要

In stochastic multi-objective bandits, Pareto optimal arms may represent different trade-offs among objectives. Balanced sampling across them is important in practical settings such as online advertising. Yet, there seems to be little consensus on how to measure fairness, to our best knowledge. Achieving fairness is also tough as the set of Pareto optimal arms is unknown, and early sampling decisions may create imbalances that are difficult to erase afterwards. We measure unfairness by the KL divergence from the uniform distribution over the true Pareto set to the empirical measure induced by the sampling trajectory. Nicely, this measure admits an exact decomposition that separates the cost of sampling suboptimal arms from the imbalance among Pareto optimal arms. With a recursive form for unfairness, we develop \emph{adaptive top-two sampling}, which combines sampling counts and empirical evidence of Pareto membership to define a gain for each arm and guide sampling without committing to an estimated Pareto set. When the Pareto set is stable under small perturbations, we establish an $O(\log T/T)$ and $O(T^{-1})$ bound on expected unfairness and variance, respectively. Information-theoretic lower bounds show that the $\log T/T$ rate can be unavoidable for uniformly fair policies on stable instances, matching our upper bound. More surprisingly, stability is indeed crucial, as the expected unfairness can be $ω(\log T/T)$ without it. Finally, we create a variant that encourages exploration of undersampled arms to reduce the risk of persistent underallocation. At sufficiently large horizons, it preserves the rate on the expected unfairness while admitting a sharper bound of $O(\log^2 T/T^2)$ on variance. Numerical studies support these findings and suggest practical insights.

cs.LG · stat.ML

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35550

Cross-Domain Sequential Recommendation with Representation-Level User Transfer

Abdulaziz Samra, Nikita Sukhorukov, Alexey Grishanov, Alexey Vasilev, Evgeny Frolov

原文摘要

Cross-domain sequential recommendation aims to improve next-item prediction in a target domain by leveraging user behavior from auxiliary source domains. However, real-world deployment faces two critical challenges: business constraints that isolate interaction data across domains, and user heterogeneity, where many users lack cross-domain activity. This makes existing cross-domain methods difficult to apply. In this paper, we propose a simple and practical framework for cross-domain sequential recommendations that transfers information exclusively through learned user representations, without sharing raw interaction data or item embeddings across domains. We introduce a lightweight adapter that projects the source-domain user representation into the target-domain embedding space and integrates it with the target interaction sequence. Importantly, when source data is unavailable (i.e., the user observed only in the target domain), the model seamlessly defaults to a robust target-only sequential recommender, retaining benefits from joint training across all users. Extensive experiments were conducted under two setups; one targets benchmarking against previous baselines, while the other focuses on realistic and scalable conditions. Both sets of experiments demonstrate consistent performance gains for shared and unshared users.

cs.IR · cs.AI · cs.DC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35549

Explainable Prediction of High-Speed Solar Wind Streams driven by Coronal Holes via Topological Data Analysis

Francesco Marchetti, Sabrina Guastavino, Daniel Collin

原文摘要

High-speed solar wind streams (HSSs) originating from coronal holes (CHs) are an important driver of recurrent geomagnetic activity. Since HSSs are associated to coronal structures visible in extreme-ultraviolet observations, this provides a natural basis for forecasting several days in advance. In this work, we investigate whether the topology of coronal structures observed in extreme ultraviolet (EUV) images contains predictive information on the subsequent occurrence of CH-driven HSSs at Earth. To this end, we consider Topological Data Analysis (TDA), and in particular persistent homology, to classify EUV solar images according to the occurrence of HSSs at Earth four days later. The target is a binary classification of HSS occurrence at a four-day lead time, defined using an HSS interval catalog and solar wind speed thresholds. Several combinations of topological representations and classifiers are evaluated under a common cross-validation framework. The results indicate that the topology of EUV coronal structures alone contains predictive information on subsequent CH-driven HSS occurrence, despite the absence of explicit CH geometric descriptors in the model input. To investigate the origin of this predictive information, we develop a model-level explainability framework based on cross-validation-aggregated feature weights and spatial evidence maps. A fully traced example connects persistence diagrams and persistence-image features to the classifier output and to their spatial localization in the original EUV observation. The proposed framework provides a transparent and computationally lightweight approach for investigating the relationship between coronal topology and subsequent high-speed solar wind streams.

astro-ph.SR · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35545

Efficient probabilistic weather forecasting at quarter-degree resolution with two-sided transport attention

Junu Choi, Youngjoon Hong

原文摘要

High-resolution probabilistic weather forecasting remains computationally demanding. Here we present Astra, a 0.25$^\circ$ ensemble model trained on eight H200 GPUs. It combines Kantorovich Attention, a two-sided transport attention in which each region chooses where it reads from and writes to, so every region sends and receives at token resolution, with a self-featurised continuous ranked probability score (CRPS) that also scores forecasts in the model's multiscale features. In WeatherBench 2, Astra matches GenCast's CRPS averaged over five upper-air targets and days 1-10 with 9.8% of its hardware-normalised training compute. Verified against ERA5, its CRPS and ensemble-mean root-mean-square error are lower than the ECMWF ensemble's on all 600 variable, level and lead-time combinations, and its CRPS is lower than GenCast's on 73.5% of them. A 15-day member needs 39 times fewer floating-point operations than GenCast's GPU implementation, and 50 members take under 40 min on one consumer GPU, within 21 GB.

cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35544

StateLoop: Persistent State for Looped Transformers

Feichen Song, Boyi Zeng, He Li, Shixiang Song, Yiqin Hao, Zitong Wang, Yizhou Yang, Yifan Liu 等 10 位作者

原文摘要

Looped Transformers offer a parameter-efficient approach to language modeling by repeatedly applying a shared backbone. Vanilla looped Transformers use each backbone output directly as the input to the next iteration. These outputs may contain information useful several iterations later, but retaining it across the loop depends entirely on the single hidden state. We introduce StateLoop, which maintains one additional persistent state per token to retain information across iterations. At each iteration, a learned gate controls how much of the previous state is retained and how much information from the backbone output is incorporated. The updated state is projected back and added to the backbone output for subsequent iterations and final prediction. This design reduces Pile perplexity and improves average accuracy across eight zero-shot tasks over Vanilla Looped. Its perplexity advantage over Vanilla Looped grows with trained loop depth, consistent with persistent state becoming increasingly useful as recurrent computation deepens.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35543

Tiny Suffix Arrays

Sharma V. Thankachan

原文摘要

Suffix array and inverse suffix array queries on a text of length $n$ can be answered in $\widetilde{O}(τ^3)$ time using an auxiliary data structure occupying $O(n/τ)$ words, for any $τ\in [1,n]$.

cs.DS · cs.IR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35540

Beyond Modeling Order: Dual-Path Spatiotemporal Representation Learning for Cross-Subject EEG Emotion Recognition

Zehui Xiao, Yi Zheng, Guixing Fu, Wenkai Lu, Lingfei He, Tao Tan, Yue Liu, Silun Wang 等 10 位作者

原文摘要

Cross-subject generalization remains a major challenge for EEG-based emotion recognition due to pronounced inter-subject variability in neural activity and emotion-related spatiotemporal patterns. Existing methods mainly alleviate inter subject differences through distribution alignment or contrastive learning, while paying limited attention to representation bias associated with fixed spatiotemporal modeling orders. We hypothesize that high-level emotion representations of the same EEG signal should remain consistent across different spatiotemporal modeling paths. Based on this hypothesis, we propose a Dual-Path Spatiotemporal Contrastive Learning (DP-STCL) framework for cross-subject EEG emotion recognition. DP-STCL employs two complementary encoders with independent parameters, namely a Temporal-to-Spatial (TS) path and a Spatial to-Temporal (ST) path, to model EEG spatiotemporal information under different integration orders. During representation learning, the same emotional stimulus is used to establish cross subject associations, while three complementary relations jointly constrain representation variations across subjects and modeling paths. This promotes stable and discriminative stimulus-related emotion representations. To further handle limited labeled target subject data, we introduce a LoRA-based few-shot adaptation strategy with low parameter overhead. Extensive cross-subject experiments on SEED, SEED-IV, SEED-V, and FACED show that DP-STCL consistently outperforms the compared methods, demonstrating its effectiveness in improving cross-subject EEG emotion recognition.

cs.LG

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2607.22433

VIREL: Route-Local Lattice Residual Compression for Exact and Error-Bounded Floating-Point Time Series

Yue Zhang, Jiatao Lin, Haopeng Chen

原文摘要

Time-series databases store floating-point columns in independent pages that are written at ingest, rewritten at compaction, and decoded on every scan. Some columns must return the stored IEEE 754 word; others accept a pointwise bound $\varepsilon$. Codecs for both contracts turn each value into an integer on a lattice, but a real page mixes decimal resolutions. Keeping one scale per block charges every coarse step for the finest value: one scale-6 reading in a 1,024-value page cuts Falcon's compression ratio by 36%. We present VIREL, route-local lattice residual compression. A value enters a lattice only after the encoder has replayed the decoder's arithmetic on it. VIREL-Exact replays the decoder's division at each value's least decimal scale, which we prove always succeeds below $2^{51}$, gives every resolution its own lane and history, and places averages on a rational lattice; the same finer value costs it 0.6%. VIREL-EB replays a fused multiply-add on an error lattice, so every admitted value decodes within $\varepsilon$. One integer coder serves both contracts. On 14 canonical streams, VIREL-Exact emits 27.6% fewer bytes than Falcon. Fast beats ALP on ratio and both speeds at once: 7.01$\times$ against 4.19$\times$, 476 against 120 MB/s, and 5.0 against 4.5 GB/s on one core. Upper is the smallest output on all 48 larger streams. Ingest with Fast and compaction with Upper store those streams in Apache TsFile at 48.7% of ALP's bytes. VIREL-EB beats Machete on ratio and both speeds in all 90 dataset-bound cells.

cs.DB

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35536

EARS: Evidence-Aware Retrieval via Sound and Motion for Budgeted Video Reasoning

Shih-Yao Lin

原文摘要

Uniform frame sampling can miss brief evidence in sparse-event videos when the VLM sees only K frames. We propose EARS (Evidence-Aware Retrieval via Sound and motion), a training-free, query-free front-end that ranks frames using optical-flow motion and audio energy, then applies temporal non-maximum suppression for diversity. The evidence score alone has frame-level AP 0.39; used only to select frames for the VLM, EARS with temporal diversity reaches localization AP 0.94, compared with 0.88 under Uniform sampling. In the updated matched study, at K = 1 EARS reaches XD-subset F1 0.660 versus 0.614 for centered Uniform (the paired interval includes zero); earlier backbone studies use separate protocols. On mostly-silent UCF-Crime, however, EARS slightly trails Uniform. A single configuration is used across both benchmarks; the new controls do not repeat the original held-out search. EARS therefore targets low-budget videos whose salient evidence aligns with motion or audio; on dense narratives or stealthy events, uniform or semantic selection remains preferable, and cheap modality statistics cannot reliably route between them. Code: https://github.com/mikeslin-ir/EARS

cs.CV

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35533

An Algorithmic Account of Few-Shot Concept Induction in Language Models

Michael A. Lepori, Laila Johnston, Aalok Sathe, Zhuonan Yang, Roman Feiman, Ellie Pavlick

原文摘要

Both models and minds not only encode features that describe an entity, but can specifically represent which features of that entity are relevant for a given context. This core capability underlies in-context learning in language models and a variety of cognitive phenomena in minds, yet it is unclear how intelligent systems identify and promote relevant features. In this work, we propose Amortized Direction Accentuation (ADA), an algorithmic hypothesis that provides an explanation for how relevance is determined in the context of few-shot concept induction --- the ability to infer a latent concept from a small number of positive examples. Under ADA, concept induction proceeds by forming a summary representation of the examples in a feature space that is shaped by amortized knowledge about which features tend to be useful for finding commonalities between items. We investigate and validate ADA in language models using a few-shot concept induction task. This work provides a novel perspective on the mechanisms supporting contextualization in language models and a promising hypothesis for a core aspect of cognition.

cs.CL

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.29924

DCGrasp: Distance-aware Controllable Grasp Generation

Hiroyasu Akada, Jesús Pérez, Emre Aksan, Vasileios Choutas, Cristian Romero, Alberto Garcia-Garcia, Vladislav Golyanik, Christian Theobalt 等 9 位作者

原文摘要

Generating 3D hand-object interactions is essential for applications in robotics, AR/VR, and synthetic data generation, where flexible controllability and strong generalization to diverse object geometries are required. However, existing methods rarely satisfy these requirements, limiting their practical applicability. We present DCGrasp, a multi-modal diffusion framework for distance-aware controllable grasp generation. DCGrasp is built on a novel grasp energy term that computes Distance Profile, a signed distance from each hand vertex to the nearest object point, coupled with distance-aware weighting, effectively capturing the semantically similar hand-object interaction in near-contact regions while remaining invariant to object and hand identity. Given various controllable signals, DCGrasp first generates a Distance Profile based on a Diffusion Transformer, together with a corresponding candidate hand pose. We then refine the candidate pose through optimization, enforcing consistency between the optimized hand pose and the generated Distance Profile in near-contact regions. Our experiments show that DCGrasp produces high-quality, geometrically plausible grasps with flexible user control, generalizing to diverse object and hand shapes and scales. Our work establishes a robust and versatile pipeline for the synthesis of controllable 3D hand-object interactions. The supplementary video with dynamic visualization is available on our project page.

cs.CV

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35531

One Loss Resolves Signed Rank-One Gradient Fibers: Exact Decoding on Cycle and Theta Blocks

Junwen Lou, Shuo Tian, Penghui Liu

原文摘要

For masked rank-one least squares with unrestricted signs, gradients reveal weighted margins and the loss reveals the target's squared norm. We prove that this one scalar resolves the gradient ambiguity on bipartite cycle and length-three parallel-path blocks: every target fixed before sampling and nonzero on cycle edges is identified almost surely by one legal positive query drawn from an absolutely continuous law. On six-cycle flowers, open target and query sets retain $3^r$ bounded gradient candidates. A common algebraic witness excludes cross-block energy cancellation. Cactus masks admit explicit exception certificates and target-uniform finite-grid bounds; theta blocks admit an existence guarantee. For nondegenerate cactus reductions, an exact decoder returns complete actual real lists, with bit cost polynomial in input length, the cycle-degree product $D$ and the optional budget parameter $2^κ$. Energy-coordinate propagation certifies singletons without a product join. The product $D$ can grow exponentially in cycle count. Exponential certificate-degree and output bounds, and weak NP-hardness at a fixed uniform query, characterize distinct costs of exact decoding.

cs.SC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35530

CANS: Capability-Preserving Neuron Steering for Large Language Models

Yucheng Wu, Chengyi Xie, Chenpeng Wang, Liangming Pan

原文摘要

Activation steering offers training-free and dynamic control of target behaviors in large language models, but existing methods can impair general capabilities and require domain-specific intervention designs. We introduce Capability-Preserving Neuron Steering (CANS), a unified framework that balances target-behavior control with capability preservation. To make target-behavior improvement and general-capability preservation tractable, we introduce two proxies: log-probabilities of target tokens in model generated responses, and output-distribution KL divergence on general-purpose inputs. Combining the local approximation of the two proxies yields a closed-form solution, enabling lightweight steering without iterative model training or additional inference-time computation. The resulting neuron-wise scaling coefficients achieve effective steering while limiting general-capability loss. A shared data-construction and intervention procedure supports adaptation across domains using sampled responses with target-behavior filtering and target-span annotations. Across safety, toxicity, and hallucination control on Qwen and Llama models, CANS achieves comparable or better steering while better preserving MMLU, BBH, and IFEval performance than SFT, CAA, and neuron-level baselines. On Qwen3-4B, CANS reduces HarmBench attack success from $15.50\%$ to $1.75\%$ with a $1.11$-point decrease in average capability. These results support CANS as a practical approach to behavioral control across domains.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35529

WorldPager: Zero-Stall Tiered Memory for Long-Horizon World Models on a Single Consumer GPU

Bingnan Liu, Ruini Xue

原文摘要

Interactive world models keep the KV of past chunks as retrieval memory so that revisited places stay consistent. On a single consumer machine (one RTX 4090, 32 GiB of host memory, one NVMe SSD), long sessions with such memory hit two walls. The first is capacity: the KV of Matrix-Game 2.0 exhausts host memory after 45 seconds of video in bf16 and after 2.8 minutes in int4. The second is the model's positional horizon of 4.3 minutes. We show that this memory is capacity-bound rather than bandwidth-bound and that its access set is known before compute, because the camera pose is an input. An analytical model built on these observations predicts spill points to within one chunk and classifies zero versus non-zero stalls correctly on held-out configurations. Guided by it, WorldPager stores int4 KV in a GPU/DRAM/NVMe hierarchy instead of recomputing it, prefetches each chunk's memory from the extrapolated pose, never waits on a miss, and writes back in the background; to our knowledge, it is the first to place the KV memory of an interactive world model on an NVMe tier with zero-stall prefetching. With training-free position handling, it generates 37.5 minutes of video continuously with zero stalls and constant host memory, at about 9.5 frames per second. On a sealed test set, int4 storage shows no measurable quality loss; on long-gap revisits it lowers image quality by about one MUSIQ point.

cs.DC · cs.CV

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2510.21978

Beyond Reasoning Gains: Mitigating General-Capability Forgetting in Large Reasoning Models

Hoang Phan, Xianjun Yang, Yuanshun Yao, Jingyu Zhang, Shengjie Bi, Xiaocheng Tang, Madian Khabsa, Lijuan Liu 等 9 位作者

原文摘要

Reinforcement learning with verifiable rewards (RLVR) has delivered impressive gains in mathematical and multimodal reasoning and has become a standard post-training paradigm for contemporary language and vision-language models. However, the RLVR recipe introduces a significant risk of capability regression, in which models forget foundational skills after prolonged training without employing regularization strategies. We empirically confirm this concern, observing that open-source reasoning models suffer performance degradation on core capabilities such as perception and faithfulness. While imposing regularization terms like KL divergence can help prevent deviation from the base model, these terms are computed on the current task and therefore do not guarantee preservation of broader knowledge. Meanwhile, commonly used experience replay across heterogeneous domains makes it nontrivial to decide how much training emphasis each objective should receive. To address this, we propose RECAP-a replay strategy with dynamic objective reweighting for general knowledge preservation. Our reweighting mechanism adapts online using short-horizon signals of convergence and instability, shifting the post-training focus away from saturated objectives and toward underperforming or volatile ones. Our method is end-to-end and readily applicable to existing RLVR pipelines without training additional models or heavy tuning. Extensive experiments on benchmarks using Qwen2.5-VL-3B and Qwen2.5-VL-7B demonstrate the effectiveness of our method, which not only preserves general capabilities but also improves reasoning by enabling more flexible trade-offs among in-task rewards.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2509.25002

Circuit Distillation

Somin Wadhwa, Silvio Amir, Byron C. Wallace

原文摘要

Model distillation typically focuses on behavioral mimicry, where a student model is trained to replicate a teacher's output while treating its internal computations as a black box. In this work we propose an alternative approach: Distilling the underlying computational mechanisms implemented by a teacher model. Specifically, we propose circuit distillation, which introduces an objective to align internal representations between analogous circuit components in teacher and student models. We propose a method to match ``functionally correspondent'' circuit components and introduce a loss reflecting similarities between the representations that these induce. We evaluate circuit distillation on entity tracking and theory of mind (ToM) tasks using models from the Llama3 family. Our results demonstrate that circuit distillation outperforms standard distillation, successfully transferring algorithmic capabilities by adjusting only a small, targeted subset of student model parameters. This work establishes the feasibility of transferring mechanisms, which may in turn allow for efficient distillation of targeted teacher capabilities via interpretable and controllable internal student mechanisms.

cs.CL

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35523

Ranking Function Generation with LLMs for Termination

Rongkang Li, Jiaxin Hu, Taolue Chen, Yong Li, Fu Song, Zhilin Wu

原文摘要

Ranking functions are a fundamental technique for proving loop termination, but constructing them manually is challenging and requires substantial expertise. We propose RAINFALL, a neuro-symbolic framework which follows the generate-verify-refine paradigm and tightly integrates LLMs with symbolic verification. RAINFALL formulates ranking function synthesis as a code-completion task by inserting placeholders into the input program and constraining the LLM to fill four predefined slots for each loop. The completed program is then checked by a symbolic verifier. If verification succeeds, a valid ranking function is obtained; otherwise, the verifier's diagnostics are translated into natural-language feedback that guides iterative refinement of the LLM's output. We evaluate RAINFALL on two public benchmark suites comprising 358 terminating C programs. Using MiniMax V2.7, RAINFALL synthesizes verifiable ranking functions for 323 programs (90.22%), outperforming all promising baselines. In particular, it solves 12 more programs than PROTON 2.1, the best-performing baseline and winner of the SV-COMP 2026 Termination category. Furthermore, combining RAINFALL with PROTON 2.1 increases the success rate to 96.65% (346/358 programs). We also conduct ablation study confirming individual contributions of placeholder contracts, few-shot examples, symbolic verification, and diagnostic information feedback. Experiments across LLMs further confirm the effectiveness of RAINFALL. Our results suggest a generic, sound framework for ranking-function synthesis and, beyond termination, LLM-assisted synthesis of formally checkable liveness certificates, both of which represent new directions of software engineering research.

cs.SE · cs.AI · cs.LO

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2412.12786

Forbidden Patterns in Mixed Linear Layouts

Deborah Haun, Laura Merker, Sergey Pupyrev

原文摘要

An ordered graph is a graph with a total order over its vertices. A linear layout of an ordered graph is a partition of the edges into sets of either non-crossing edges, called stacks, or non-nesting edges, called queues. The stack (queue) number of an ordered graph is the minimum number of required stacks (queues). Mixed linear layouts combine these layouts by allowing each set of edges to form either a stack or a queue. The minimum number of stacks plus queues is called the mixed page number. It is well known that ordered graphs with small stack number are characterized, up to a function, by the absence of large twists (that is, pairwise crossing edges). Similarly, ordered graphs with small queue number are characterized by the absence of large rainbows (that is, pairwise nesting edges). However, no such characterization via forbidden patterns is known for mixed linear layouts. We address this gap by introducing patterns similar to twists and rainbows, which we call thick patterns; such patterns allow a characterization, again up to a function, of mixed linear layouts of bounded-degree graphs. That is, we show that a family of ordered graphs with bounded maximum degree has bounded mixed page number if and only if the size of the largest thick pattern is bounded. In addition, we investigate an exact characterization of ordered graphs whose mixed page number equals a fixed integer $ k $ via a finite set of forbidden patterns. We show that for every $ k \ge 2 $, there is no such characterization, which supports the nature of our first result.

cs.DM · math.CO

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35522

Generative Structure from Motion with Native 3D Diffusion

Haobin Duan, Binbin Huang, Mochu Xiang, Yiqun Zhao, Zibo Zhao, Yi Ma, Shenghua Gao

原文摘要

Recovering an object's 3D structure and camera motion from a few uncalibrated views is a fundamental problem in computer vision. Classical structure from motion methods recover 3D geometry consistent with the input images but cannot recover unseen regions; native generation methods produce plausible 3D objects which are related but not necessarily faithful to the input images. Methods simply concatenating reconstruction and generation allow error to cascade across the two stages, instead of producing a 3D object consistent with both the distribution and the input. We propose GenSfM, which integrates image-based 3D reconstruction with a native 3D diffusion model. In particular, we use a multi-modal diffusion transformer to jointly denoise the shape latent and per-view pose latents under shared self-attention, grounding generated geometry in the observed views while using object-level priors to infer unobserved regions. A compact 3D-native UV-volume tokenization keeps the joint sequence small as views are added, and a pose-aware stage refines geometry and texture by sampling per-voxel image features at the recovered camera poses. On Toys4k and GSO, for $4$ uncalibrated views, GenSfM improves novel-view PSNR by $3.2$-$4.5$ dB and reduces median Chamfer distance by $2.8$-$3.8\times$ over a strong native-3D generation baseline, while recovering input-view poses over $96\%$ Acc@$5^\circ$ without any external pose-estimation backbone, post-hoc optimization, or known intrinsics. Project page: https://dotb1ue.github.io/GenSfm/.

cs.CV · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35521

PRISMA: Differentially Private Synthetic In-Car Telemetry via Differentiable Simulation

Kilian Zeiseweis, Nele Sophie Brügge, Marven Kummerfeldt, Ali Raza, Esfandiar Mohammadi

原文摘要

Differentially private synthesis of in-car energy consumption data is challenging because the output is high-dimensional, and known lower bounds show that the privacy-utility trade-off degrades with the square root of the number of output dimensions. However, we observe that the intrinsic dimensionality of in-car energy consumption is much lower than that of the raw trace: given a small set of contextual personal variables-such as start point, end point, the car model, and the driving behavior-the consumption profile can be largely explained by publicly available map data (including elevation) combined with a physical model of vehicle energy use. We exploit this low intrinsic dimensionality in PRISMA, a Privacy-preserving In-car Synthesis and Modelling Algorithm. PRISMA involves three stages: (i) energy consumption traces are encoded into a low-dimensional latent space; (ii) the distribution of latent vectors is synthesized using the differentially private tabular data synthesis algorithm AIM, and (iii) samples drawn from this distribution are decoded back into full energy consumption traces via a differentiable physics-based simulation, a step that consumes no additional privacy budget since it uses only public information. By using a compact, physically-grounded latent representation of the traces, PRISMA significantly reduces the effective output dimension. Our experimental results show that PRISMA achieves for a competitive $\varepsilon = 0.5$ a better data synthesis performance than prior tabular and marginal-based DP synthesis baselines by preserving the diversity of the original power consumption profiles without collapsing to a small set of modes.

cs.CR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.35518

S3AC: Stabilized, Selective, and Scalable Actor-Critic for LLM Reasoning

Shuo Shen, Guangran Cheng, Chengqi Lyu, Xiangsen Wang, Yanan Sun, Wenwei Zhang

原文摘要

Critic-free methods are widely used in reinforcement learning with verifiable rewards (RLVR) for long chain-of-thought (CoT) reasoning, avoiding difficult and costly critic training. Common alternatives lack token-level credit assignment and typically rely on grouped rollouts to estimate prompt-specific baselines. A critic can address both limitations, but long-CoT critic training is hindered by insufficient coverage of trajectories with diverse lengths and early gradients dominated by the newly initialized value head. We introduce S3AC (Stabilized, Selective, and Scalable Actor-Critic) for LLM reasoning, built around a practical critic-training recipe. The trained critic provides token-specific advantages and prompt-specific baselines. Using the prompt-level baseline to estimate difficulty, we propose Difficulty-Aware Prompt Sampling (DAPS). DAPS combines this estimate with outcome history to select prompts before generation, reducing rollout waste without offline difficulty initialization. S3AC improves group-rollout performance across dense and mixture-of-experts models from 1.5B to 397B parameters, outperforming a modern GRPO variant (GRPO-X). At 35B, it sustains single-rollout training while tested critic-free baselines deteriorate after initial gains.

cs.LG · cs.CL

查看论文
arXiv ↗PDF ↗