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

+479新增论文
+361其中 AI 相关
+63已有论文修订
+93新增 AI 导读

arXiv 最近完成扫描:2026-11-16 01:34 · 全站最近入库:2026-11-16 13:44各信源进度 →

统计口径与数据新鲜度

统计 2026-11-15 21:35 — 2026-11-16 21:35(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:2026-11-16 01:34最近发现新增版本:2026-11-16 13:44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。

23,827 已收录论文19,941 AI 相关6,921 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 23,827 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-16 01:33(北京时间)
新论文 · v1arXiv:2611.11893

Developing a Street View Imagery Processing Pipeline for Building-Level Vacant House Assessment

Dongjoon Park, Sunjae Lee, Jung Woo Kim, Bumjoon Kang

原文摘要

Street view imagery (SVI) is widely used to audit urban environments remotely. Automated assessments usually assume that each image shows the target building clearly. In practice, an image may lack task-relevant evidence or show a neighboring house. If this goes unchecked, a model's score cannot show whether it judged the correct building. We present a reproducible SVI processing pipeline that makes image suitability and target identity explicit and checkable. We apply it to vacant house assessment in South Korea, using 10,311 images of 1,609 detached-house parcels. Task-suitability filtering adapts its criteria to image composition and administrative area. This keeps coverage even across areas. While keeping as many images as a single uniform filter, our filtering left 101 more parcels with a usable image. Target-house alignment identifies the parcel's house from camera metadata and parcel boundaries. It needs no building footprints or additional training. Alignment matched human labels for 85.6% of images, compared with 79.8% for the best simple rule. In vacancy classification, filtering and alignment raised AUROC from 0.684 to 0.719, but marking a different house in the same image still gave 0.709. Grad-CAM showed that the model often looked at areas outside the target house. Image suitability, target identification, and model evidence are therefore separate problems, and scores alone do not resolve them. Recording which images were used and which house was selected lets reviewers verify each assessment and can support human virtual audits.

cs.CV · cs.CY

新论文 · v1arXiv:2611.13017

Multi-Future Drafting: Mixture Block Drafters for Tree-Verified Speculative Decoding

Haoyu Huang, Jialiang Kang, Linlin Yang, Xiao Yang, Baochang Zhang

原文摘要

Block drafters such as DFlash draft a block of future tokens in one forward pass, and draft trees such as DDTree verify the prefixes ranked highest by the product of the drafter's per-position distributions. Under tree verification, the expected number of accepted tokens is the probability mass the tree covers. We show that a product misallocates it: when the continuation can follow several futures, it splices them into chimeras, prefixes stitched together from different futures that the target never produces, and gives them as much budget as the futures themselves. Barring ties, no product-ranked tree can hold two futures that disagree in $m$ positions without $2^{m-1}-1$ chimeras, however well the drafter is trained, and in practice chimeras make up 27-33% of a 64-node tree. We propose Multi-Future Drafting (MFD), which restores the coupling with a latent variable instead of the chain rule: $K$ learned mode embeddings at the drafter's input draft $K$ coherent blocks in the same pass, and a mixture DDTree lets the verifier pick the future the target takes, losslessly and with $K+1$ extra vectors. On four Qwen targets from 4B to 27B parameters, at batch size 1, MFD reaches up to 7.8$\times$ speedup, 5-8% faster than the best single-mode tree trained identically and, on Qwen3-4B and Qwen3-8B, 9-16% faster than the released DARTree trained on the same prompts.

cs.AI · cs.LG

新论文 · v1arXiv:2611.11892

When Answer Decodability Is Not a Decision: Separating Final-Answer and Imminent-Emission Signals in Chain-of-Thought

Hyojung Han, Jongmin Kim, Seung-Hun Jeon

原文摘要

Answer probes applied to chain-of-thought hidden states are often read as evidence that a language model has already decided its final response. This reading overlooks a competing signal, the candidate the model is about to mention. We introduce emission-controlled dual-target probing (ECDP), which fits the same low-rank reader family separately to the imminent mention and to the final answer at identical positions, with target-specific nulls and thresholds fixed before any fit. On correct DecisionDepth (DD) rollouts in which a distractor is mentioned before the final answer, Qwen3.5-4B decodes the next mention at .966 (null .234) but the final answer only at .221 (null .186), an emission gap of .696 [.648, .744]; Llama-3.1-8B gives .575 [.514, .632]. This pattern is not universal. In a pre-specified evaluation of Qwen3-8B with thinking enabled on held-out MMLU questions, using a full-prefix attention reader over every thinking position, both targets are decodable and the final answer is read far more strongly (mention-minus-answer lift -.398 [-.459, -.336] at layer 22 over 467 questions). Model, task, positions, and reader change together between these settings, and an attempted matched comparison on DD produced no valid statistics because the imminent-mention probe did not train reliably, so we do not identify what reverses the order. On DD, planted-signal controls show that the reader recovers an injected final-answer signal at one tenth of full dose. Answer-candidate decodability is therefore target-relative and protocol-dependent: a strong answer readout does not by itself establish when a model commits, and a probe that fails to train is not evidence that the information is absent. ECDP is a practical control for distinguishing imminent text emission from evidence of a settled decision.

cs.LG · cs.CL

新论文 · v1arXiv:2611.11891

A Large-Scale Dataset for Cross-Lingual Multi-Jurisdiction Patent Drafting from Patent Cooperation Treaty Applications

Trung Phuong Le, Hirofumi Nonaka, Koji Marusaki, Katsuhito Sudoh, Seiya Kawano

原文摘要

Securing patent protection across countries requires drafting claims separately for each national patent office, in the relevant language and according to its legal and stylistic conventions. Patent claims can therefore differ across languages and jurisdictions because of translation, target-office drafting practices, and prosecution-stage changes. We formulate cross-lingual patent drafting as generating plausible candidate claim sets for a specified target language, jurisdiction, and prosecution stage from a source claim set in another language. The task models jurisdiction- and stage-specific transformations beyond content-preserving translation. We construct a large-scale Japanese-English dataset by linking Japan Patent Office (JPO) and United States Patent and Trademark Office (USPTO) records through 884,179 Patent Cooperation Treaty (PCT)-derived patent families. Across six released configurations, these families yield 4,067,168 directional claim-set instances; four core configurations form the main benchmark. Commercial machine translation and zero- and few-shot prompting capture part of the task, while supervised adaptation provides the strongest overall baseline, showing that task-specific patterns can be learned from linked cross-jurisdictional claim sets.

cs.CL · cs.AI

新论文 · v1arXiv:2611.11890

Training Vision Transformers to Focus: Emergence of Spatial Attention Head Specialization

Yuji Nozawa, Yu-Chieh Lin, Youyang Ng

原文摘要

Self-supervised learning in Vision Transformer models, such as DINO, enables robust visual perception without requiring labeled data by leveraging large-scale image datasets. However, these models lack a focusing mechanism during training, resulting in perceptual representations that are often ambiguous or overly similar. In this work, we introduce two visual focusing loss functions designed to establish correspondence between the model's attention and specific regions within images. Specifically, we leverage the multi-head attention mechanism within the model to selectively steer the model's focus, leading to the emergence of spatial attention head specialization, characterized by diverse and spatially focused perceptual capabilities without requiring annotations. Through both qualitative and quantitative evaluations, we demonstrate that our method substantially increases the spatial selectivity and diversity of attention heads, improving both interpretability and performance on dense prediction tasks such as zero-shot image segmentation.

cs.CV

版本更新 · v2arXiv:2606.29129

Improved Scaling for Fast Mode of Ozaki Scheme II

Shota Kawakami, Daisuke Takahashi

原文摘要

Ozaki scheme II emulates high-precision matrix multiplication using low-precision integer matrix operations based on the Chinese remainder theorem (CRT). It first scales the high-precision matrices to convert them into integer matrices. For this scaling step, Ozaki scheme II provides two modes: accurate mode, which uses INT8 matrix multiplication to estimate scaling factors, and fast mode, which applies the Cauchy--Schwarz inequality at lower computational cost. We show that the existing formula lacks scale invariance; multiplying the input matrices by a constant changes the effective bit width of the integer matrices in the scaling step, causing accuracy degradation or CRT recovery failure. To address this, we propose a revised scaling formula derived from the CRT uniqueness condition via the Cauchy--Schwarz inequality. The proposed formula is scale-invariant by construction, guarantees that the CRT uniqueness condition is always satisfied, and introduces no additional overhead over the original fast mode. Experiments on an NVIDIA GH200 GPU show that the proposed method achieves accuracy comparable to that of accurate mode while maintaining throughput comparable to that of fast mode. In the accuracy--throughput trade-off, the proposed method overcomes the accuracy limitation of fast mode and the throughput constraint of accurate mode, offering a superior accuracy and performance.

cs.MS · cs.DC · math.NA

版本更新 · v2arXiv:2602.07732

Efficient Adaptive Data Analysis over Dense Distributions

Joon Suk Huh

原文摘要

Modern data workflows are inherently adaptive, repeatedly querying the same dataset to refine and validate sequential decisions, but such adaptivity can lead to overfitting and invalid statistical inference. Adaptive Data Analysis (ADA) mechanisms address this challenge; however, there is a fundamental tension between computational efficiency and sample complexity. For $T$ rounds of adaptive analysis, computationally efficient algorithms typically incur suboptimal $O(\sqrt{T})$ sample complexity, whereas for arbitrary distributions, algorithms with $O(\log T)$ sample complexity are computationally intractable under standard cryptographic assumptions. In this work, we shed light on this trade-off by identifying a natural class of data distributions under which both computational efficiency and logarithmic sample complexity in $T$ are achievable. We propose a computationally efficient ADA mechanism that attains $O(\log T)$ sample complexity, optimal in its dependence on $T$, when the data distribution is dense with respect to a known prior. This setting includes, in particular, feature-label data distributions arising in distribution-specific learning. As a consequence, our mechanism also yields a sample-efficient (i.e., $O(\log T)$ samples) statistical query oracle in the distribution-specific setting. Moreover, although our algorithm is not based on differential privacy, it satisfies a relaxed privacy notion known as Predicate Singling Out (PSO) security (Cohen and Nissim 2020) whenever the number of its possible update transcripts is polynomial in the sample size. Our results thus reveal an inherent connection between adaptive data analysis and privacy beyond differential privacy.

cs.LG · cs.DS

新论文 · v1arXiv:2611.11888

OmniSynth: A Meta-Evaluation Framework for Benchmarking the Quality of Missingness-Inclusive Synthetic Health Data

Annabel Zhu, Océane Doremus, Ariel Guerra-Adames, Marta Avalos-Fernandez, Emmanuel Lagarde, Cédric Gil-Jardiné, Vianney Jouhet

原文摘要

Synthetic data are an emerging solution to the reproducibility crisis in machine learning for health, driven by data privacy restrictions surrounding electronic health records. Yet no consensus exists on how to evaluate synthetic health data quality, existing metrics functionally overlook missingness despite its prevalence in clinical data, and benchmarking frameworks have never been empirically validated. We introduce OmniSynth, a modular Python benchmarking suite that evaluates synthetic tabular data across fidelity, missingness similarity, utility, and privacy - the first framework to explicitly evaluate missingness as a distinct dimension of synthetic data fidelity. OmniSynth further proposes a meta-evaluation procedure that stress-tests evaluation performance under controlled degradation scenarios, providing a systematic characterization of how an evaluation framework behaves across datasets, sample sizes, and variable types. Applying this procedure to OmniSynth in two health datasets (Pima Indians Diabetes; MIMIC-IV-ED), we found that scores responded monotonically to degradation, three of four novel missingness metrics captured induced missing data patterns, and individual metric contributions to the scoring system were empirically justified.

cs.DB

版本更新 · v4arXiv:2511.19842

Strategy-robust Online Learning in Contextual Pricing

Joon Suk Huh, Kirthevasan Kandasamy

原文摘要

Learning effective pricing strategies is crucial in digital marketplaces, especially when buyers' valuations are unknown and must be inferred through interaction. We study online contextual pricing with sealed bids from multiple recurring buyers. The seller selects an affine pricing policy before observing the current context and competes with the best fixed policy in the same class. Buyers may misreport their values to influence future prices, a challenge known as strategic overfitting (Amin et. al., 2013). We introduce a strategy-robust notion of regret for multi-buyer online environments, capturing worst-case strategic behavior in the spirit of the Price of Anarchy. Our first contribution is a polynomial-time approximation scheme (PTAS) for learning affine pricing policies against arbitrary context-value sequences: for every fixed $ε>0$, it runs in polynomial time and its regret is $O(εT)$ plus a term sublinear in $T$. The scheme is based on an online sketch constructed from past contexts. We then propose the Sparse Update Mechanism (SUM), which learns only from a privately selected subset of rounds and occasionally posts a random price drawn from a finite grid. SUM is a PTAS for strategic regret, and its guarantee holds uniformly over all Nash equilibria. Both schemes use an online expert algorithm as a black box and run in $(T/ε)^{O(ε^{-2})}\mathsf{poly}(d,T,1/ε)$ time. Furthermore, we show that, even with truthful bids, no seller algorithm can achieve regret $\mathsf{poly}(d,1/ε)\,T^z+εT$ with $z<1$ in time $\mathsf{poly}(d,T,1/ε)$ unless $\mathrm{NP}\subseteq\mathrm{BPP}$.

cs.GT

新论文 · v1arXiv:2611.11887

How Many Candidates Does a One-Step World Model Need? A Random-Codebook Account of Nearest-of-m Training on Frozen DINO Tokens

Yong Hwi Song, Keon Myung Lee

原文摘要

One-step stochastic world models avoid the mode averaging of deterministic feature predictors without the tens of network evaluations per step that diffusion and flow matching require. The simplest such objective is nearest-of-m matching (conditional implicit maximum likelihood estimation, IMLE): draw m candidate futures and pull only the one closest to the observed outcome. Recent work conjectures that a small m suffices for feature-space world models and that a strictly proper energy score should replace the hard minimum. We test both claims with one-step world models on frozen DINOv2/v3 patch tokens, using benchmarks whose futures can be enumerated. Under mode-like uncertainty, nearest-of-m with m=16 covers the enumerated futures better than 50-step flow matching (at about 1/30 of its latency), one-step flow generators and an energy-score generator at matched parameter counts, and its rollouts stay on the token manifold where flow rollouts drift. One quantity governs the choice of m: the nearest-of-m objective is the expected distortion of a random codebook of m draws from the generator. This view yields three predictions, recorded before the corresponding runs: (i) memorization of the realized outcome in contexts that never repeat, whose strength is governed by the codebook's miss distortion I(m) rather than its miss probability, and which an energy-score generator does not escape by raising m; (ii) a flattening of mode probabilities with m, matching a known optimum within 0.01; and (iii) under-dispersion on continuous high-dimensional outcomes, which bounds the method on real robot video. Per-step matching on self-rollouts cuts long-horizon error by a third on a stochastic Push-T variant; an ablation separates its manifold and coverage benefits. Planning success does not improve despite accurate risk estimates, a negative result we report with its diagnosis.

cs.LG · cs.CV · cs.RO

版本更新 · v3arXiv:2611.06918

From Pixels to Actions--Towards Native Vision-Language-Action Policies

Weichen Fan, Haiwen Diao, Penghao Wu, Siyuan Hu, Ziqi Huang, Tianqi Liu, Mingze Sun, Ziyi Wang 等 17 位作者

原文摘要

Vision-language-action models are typically built on a simple premise: perception is first compressed by visual encoders, and action reasoning begins only afterwards. Yet manipulation often hinges on precisely the visual signals most vulnerable to this abstraction, e.g., local geometry, boundaries, depth, and subtle state changes. This raises a fundamental question: should action policies inherit a fixed visual endpoint, or directly access evolving visual computation? We introduce NEO-VLA, a structurally encoder-free vision-language-action framework. Images are tokenized by lightweight patch convolutional layers and evolve jointly with language throughout the backbone, allowing a flow-matching action expert to draw from instruction-conditioned visual representations at different depths. This simple change reveals a consistent pattern. Native architectures outperform encoder-based counterparts under controlled comparisons; full fine-tuning reduces but does not erase the gap, and additional scale alone cannot recover it. More strikingly, layer-wise probing shows that the action-critical signals often peak well before the final visual representation. Across diverse manipulation settings and distribution shifts, NEO-VLA consequently achieves stronger and more robust control. Together, these results suggest that embodied control benefits from access to visual representations across the computational hierarchy, rather than a fixed encoder output.

cs.CV

新论文 · v1arXiv:2611.11886

Multi-Modal Agricultural Document Understanding at the DAGRI Task

Zhefan Wang, Da Li, Keping Bi, Nanqing Dong

原文摘要

Agricultural documents, such as prefectural technical reports and crop calendars, are critical knowledge resources. However, their complex table layouts, significant variations in document length, and dense multi-modal elements pose substantial challenges for automated understanding. To address these issues, the NTCIR-19 DAGRI (Data Analytics for aGRicultural Information) task introduced three subtasks focusing on different dimensions of document intelligence. In the Table Information Extraction (TableIE) task, we developed a multi-stage pipeline combining structure-preserving parsing via MinerU, region-adaptive segmentation for long documents, and a dual-modal (vision-text) extraction strategy for complex work schedules. This approach achieved first place with an exact match score of 0.6566. For Multi-Table Multi-Hop Question Answering (MTMH-QA) over interrelated JSON tables, we adopted an agentic search framework utilizing DCI-Agent to iteratively explore directories via bash commands, reinforced by a cross-model validation mechanism. This yielded a revised accuracy of 0.8444, initially ranked first. In the Figure and Table Question Answering (FiT-QA) task, which requires comprehending single large composite images, we implemented a dual-modal VLM pipeline enhanced by deterministic answer normalization, securing the highest BLEU score. Together, our system effectively addresses regional structural variations, dynamic cross-file evidence discovery, and multi-modal alignment, providing a robust framework for complex agricultural document understanding.

cs.IR

版本更新 · v2arXiv:2605.06879

Better Protein Function Prediction by Modeling Survivorship Bias

Zhongmou Chao, Poompol Buathong, Ekaterina Selivanovitch, Susan Daniel, Peter I. Frazier

原文摘要

Predicting whether a protein sequence possesses a biochemical property of interest is important for understanding and designing protein functionality. Modern protein sequence databases provide valuable training data for functional prediction, but they exhibit survivorship bias: natural selection favors variants that support organismal survival while eliminating many others before they can be observed, complicating classifier training. This setting motivates positive-unlabeled (PU) learning, which treats observed functional sequences as positive and unobserved sequences as unlabeled rather than nonfunctional. However, existing protein PU methods do not account for cross-sequence heterogeneity in evolutionary accessibility: sequences one nucleotide mutation away from a common variant arise far more readily than distant ones. We introduce PULSE: PU Learning with Surveillance and Evolution, a framework that trains a probabilistic classifier for a target protein function using an evolution-aware observation likelihood. When considering functionality of proteins with high genomic sequence coverage, PULSE combines a probabilistic classifier that estimates whether an amino acid sequence performs the target function with an evolutionary model that estimates the likelihood that its encoding nucleotide sequences emerge and are detected based on mutational accessibility, parental prevalence, and surveillance e!ciency. By coupling these two components, PULSE treats the absence of an evolutionarily accessible sequence as stronger evidence of non-functionality than the absence of one unlikely to arise. Tested on eleven tasks focusing on multiple Spike protein properties spanning five SARS-CoV-2 variants, PULSE ranks first or second on all of them, and achieves the best overall performance among all methods, including standard PU learning, one-class classification methods, and protein language models.

cs.LG · q-bio.QM

新论文 · v1arXiv:2611.11885

Design of a hybrid tendon-linkage finger with coupled and adaptive motion

Hyojae Kang, Giup Seo, Jae-Young Lee, Yeong Jae Park, Jongwoo Park, Hyunmin Do, So-Hee Park, Dongil Park 等 9 位作者

原文摘要

Dexterous robotic fingers require coordinated joint motion and adaptability to object contact, while independently actuating multiple joints increases the complexity of the transmission mechanism. This paper presents a hybrid tendon--linkage robotic finger with four mechanical degrees of freedom driven by three actuators. The MCP abduction/adduction joint is driven through a bevel gear, while MCP and PIP flexion/extension are driven by separate bidirectional tendon transmissions. The DIP joint is not independently actuated but is coupled to the PIP joint through a four-bar linkage with a torsional spring and mechanical stopper. The tendon routing also introduces mechanical coupling between the MCP and PIP joints. Kinematic analysis was conducted to describe the coupled and adaptive motions, and experiments were performed to evaluate the joint motion characteristics. ArUco-based measurements showed that the coupled motion closely matched the theoretical prediction, whereas the adaptive motion was partially limited by interference between linkage components. The fabricated finger demonstrated different contact postures with cylindrical, cuboid, and irregularly shaped objects, and the mean peak fingertip force was 4.56 N. The proposed mechanism provides both coupled motion and contact-dependent adaptation without requiring an additional actuator for the DIP joint.

cs.RO

新论文 · v1arXiv:2611.11881

When Does Physics Beat Data? Data-Efficient Traffic State Estimation with Amortized PINNs

Anna Lazareva, Alexander Tarakanov

原文摘要

Traffic state estimation (TSE) reconstructs traffic density at unmonitored locations from a sparse set of sensors. Data-driven models work well when sensors are dense but degrade sharply as coverage thins. We show that imposing the Lighthill-Whitham-Richards conservation law as a physics-informed prior makes TSE more data-efficient. On four real corridors (NGSIM, I-80, PeMS04, PeMS08), physics-informed estimators generalise to held-out sensors where data-driven models, including classical and deep kriging, overfit the observed ones. On the two shock-rich corridors the effect is a clear crossover: as more sensors are held out, data-driven models collapse while the physics-informed estimator degrades gracefully, leading the no-physics network by about 26-38% at 85-90% held out and the strongest classical kriging by about 10-21% at 90%. A matched smoothness prior does not reproduce the gain, and an extended Kalman filter on the same LWR model is outperformed on the shock corridors, so the advantage comes from how the conservation law is imposed. Finally, one loss-conditional PINN removes the per-corridor loss-weight search, matching a 20-point search at about 5x lower cost.

cs.LG

版本更新 · v2arXiv:2607.11070

MJ: Multi-turn LLM Jailbreaking via Decomposed Credit Assignment

Junyoung Park, Namgyu Park, Sechan Lee, Yoon-Chan Jhi, Jihoon Cho, Sangdon Park

原文摘要

Modern large language models (LLMs) operate in interactive multi-turn settings, making multi-turn jailbreaking a realistic threat model and an important setting for automated red teaming. A core challenge in learning multi-turn jailbreak attackers is credit assignment: different turns contribute differently to the final outcome, yet existing learning signals are often too coarse to identify their individual contributions. We propose decomposed credit GRPO (DC-GRPO), a unified turn-level credit assignment framework for Group Relative Policy Optimization in multi-turn jailbreak learning. DC-GRPO assigns a separate group-relative learning signal to each turn by combining immediate and future credit, avoiding the credit misassignment induced by broadcasting a single trajectory-level score across the dialogue. We instantiate this framework with static and dynamic weighting rules that differ in how the two credit sources are balanced while sharing the same turn-level structure. Across four victim LLMs and three benchmarks, DC-GRPO achieves up to 98.26% average ASR_5@3, substantially outperforming the state-of-the-art methods, including SEMA (86.58%) and TROJail (86.23%), demonstrating the effectiveness of our turn-level group-relative credit assignment. Warning: This paper contains examples of harmful content.

cs.CL

版本更新 · v2arXiv:2605.19235

GAE Falls Short in Imperfect-Information Self-Play Reinforcement Learning

Zhiyuan Fan, Gabriele Farina

原文摘要

Competitive multi-agent reinforcement learning in imperfect-information games requires agents to act under partial observability and against adversarial opponents, necessitating stochastic policies. While self-play reinforcement learning with Proximal Policy Optimization (PPO) has achieved strong empirical success, its standard advantage estimator, generalized advantage estimation, suffers from additional variance due to the sampling of stochastic future actions. This variance is amplified in equilibrium self-play because of the stochastic nature of the equilibrium policy and persists even when the critic is exact. We address this bottleneck by introducing $Q$-boosting, a variance-reduced advantage estimator based on a centralized action-value critic, and propose Variance-Reduced Policy Optimization (VRPO), incorporating this new estimator. The algorithm replaces sampled multi-step backups with a multi-step Expected SARSA$(λ)$ trace, computing policy expectations at each step to average out action-sampling noise, while retaining PPO's clipped objective and on-policy actor updates. Empirically, VRPO achieves strong performance in various games: In mid-sized games where exploitability is computable, VRPO lowers the exact exploitability by more than 15% compared to tuned baselines. In Dou Dizhu, it sets a new state of the art over PerfectDou. In Heads-Up No-Limit Texas Hold'em, it posts a positive win rate against Slumbot without any blueprint raise sizes, search, or subgame solving at test time.

cs.LG · cs.GT

新论文 · v1arXiv:2611.11877

Tropical Gaussian Anticoncentration: Optimal Instance-Dependent Bounds for Online Learning in Extensive-Form Games

Ashkan Soleymani, Zhiyuan Fan, Lillian J. Ratliff, Patrick Jaillet, Gabriele Farina

原文摘要

We study the minimax regret of full-information online decision making in extensive-form games. For prediction with expert advice, the optimal regret is $Θ(\sqrt{T\log K})$, where $K$ is the number of experts, or equivalently, the number of normal-form actions. For treeplex strategy spaces, existing efficient algorithms, including KOMWU and online mirror descent with the weight-one dilated entropy regularizer, achieve regret $\mathcal{O}(\sqrt{T\log|\mathcal{V}|})$, where $\mathcal{V}$ is the set of reduced normal-form strategies. We prove that this logarithmic dependence is minimax optimal in the worst-case full-information reward model. The main difficulty is that pure plans in a treeplex are not independent experts: their rewards are correlated through the recursive structure of decisions and observations. To capture this correlation, we introduce tropical Gaussian formulae, recursive Gaussian expressions built from maxima and normalized sums which preserve the variance. The index count of such a formula matches the number of reduced normal-form continuation plans in the corresponding subtree. Our main analytic result shows that every log-balanced tropical Gaussian formula with index count $N$ has Gaussian width $Ω(\sqrt{\log N})$. While a direct approach struggles to recover this lower bound, our proof uses a spherical entropy profile to measure scale-dependent separation among the Gaussian directions, and then applies Turán's theorem and Sudakov minoration to obtain the width lower bound. Applying this result to treeplexes gives an oblivious hard distribution with valid transition outcome and Rademacher terminal rewards, establishing a regret lower bound of $Ω(\sqrt{T\log|\mathcal{V}|})$. This matches known upper bounds up to universal constants and shows that the DilEnt/KOMWU dependence on the reduced normal-form complexity is unavoidable.

cs.LG · cs.GT

新论文 · v1arXiv:2611.11876

Accelerating Direct Sparse Solvers with Graph Computing Accelerators

Jiya Su, François-Henry Rouet, Jeff Beisheim, Robert Lucas, Yuqing Wang, Alexander Fell, Chris Mccarthy, David F. Gleich 等 9 位作者

原文摘要

Solving large sparse linear systems by Gaussian elimination is a challenging computational task due to the introduction of non-zero elements, or fill-in. Practitioners use heuristics such as the nested dissection reordering algorithm to try to reduce fill-in. In practice, reordering codes do not achieve the same parallel scaling as factorization, nor can they exploit AVX instructions or GPUs. Thus, reordering can be the primary runtime bottleneck in large-scale computer-aided engineering applications. We consider the use of a novel accelerator, the UpDown system, to accelerate the METIS multilevel nested dissection reordering algorithm. The UpDown system is designed for a broad class of graph computing and irregular algorithms. We report results for small-scale UpDown systems that will be available commercially in 2027, using validated simulation and FPGA emulation platforms. Our findings indicate that small UpDown systems can speed up METIS by two-orders of magnitude versus a single CPU. Compared to ParMETIS on hundreds of MPI ranks, we anticipate the same small UpDown systems offer speedups of 4 - 7, providing the potential to significantly speed finite element simulations

cs.DC

新论文 · v1arXiv:2611.11875

Beyond Head Importance: Collective Computation in Video Transformers

Yi Yang, Siyuan Liu, Bingbing Nie

原文摘要

Attention heads in video Transformers vary greatly in functional importance. Understanding their contributions requires examining not only which heads are important, but also how they participate in collective computation within the network. We study this collective computation from two perspectives: the downstream use of head outputs and the composition of interactions across layers. Ablation analyses of four video Transformers show that high-importance heads identified on a single task also support other video understanding tasks and help preserve the representational structure that distinguishes different videos. Representational responses to head ablation correlate weakly with head importance after the attention module, but more strongly after the MLP in the same block. Adapting this downstream processing through MLP-only fine-tuning further improves the use of lower-importance head outputs. Meanwhile, through composition across layers, sets of heads with shorter spatiotemporal attention distances support stronger perturbation responses at distant tokens than sets with longer distances. Their joint ablation causes even greater performance loss than ablating equally sized sets of high-importance heads, revealing collective contributions not fully captured by individual importance rankings. These findings highlight collective computation as a key perspective for understanding attention head function: functional contributions depend on both how head outputs are used downstream and how local interactions compose across layers.

cs.CV

版本更新 · v2arXiv:2608.01298

UDT: Reconciling U-Nets and Diffusion Transformers with Data-Adaptive Token Reduction

Junno Yun, Yaşar Utku Alçalar, Mehmet Akçakaya

原文摘要

Diffusion Transformers (DiTs) have emerged as a core architecture in generative modeling due to their scalability and adaptability to multimodal tasks. DiTs comprise isotropic transformer blocks, and learn representations progressively across depth, where the denoising objective drives later layers to focus on fine-detail reconstruction. This results in degraded representation quality and an imbalanced encoder-decoder behavior. Prior approaches such as representation alignment (REPA) mitigate this by encouraging stronger early representations via training regularization. Alternatively, U-Net-style DiT architectures introduce explicit multi-scale encoder-decoder structures for improved convergence. But they build on standard U-Net wisdom via learnable operators for spatial downsampling, which are not well-suited to transformer architectures, introducing inefficiencies and compatibility issues with components such as cross-attention and representation regularization. In this work, we propose UDT, a U-Net diffusion transformer that combines the representation power of DiTs with the encoding-decoding benefits of U-Nets, through data-adaptive token merging for downsampling and upsampling, while preserving the DiT token dimension. Our baseline UDT architecture outperforms existing U-Net DiTs and achieves performance comparable to REPA across all model sizes. Furthermore, using architectural optimization and REPA, UDT outperforms SiT's 7.9 FID at 1400 epochs (w/o CFG) within 40 epochs (~ 40x faster convergence) for XL model size on 256x256 ImageNet. Finally, it achieves strong image generation performance with CFG, reaching FID of 1.38 (320 epochs) with SD-VAE and 1.35 (500 epochs) with VA-VAE, providing a new backbone for DiTs with strong empirical benefits. Code is available at https://github.com/JN-Yun/UDT

cs.CV · cs.AI · cs.LG

新论文 · v1arXiv:2611.11873

Where Does Proposal Selection Fail? Diagnosing Frozen Vision-Language Queries over Detector Proposals

Bo Ma, Weiqi Yan, Jinsong Wu

原文摘要

In modular grounding a frozen detector proposes regions and a vision-language model (VLM) selects one, at lower cost than letting the VLM decode a box, but it is unclear where such systems fail. We hold perception fixed---every trained selector chooses among the same frozen proposals, which contain the referent for at least $99.8\%$ of RefCOCO/+/g queries---and vary the selection interface one factor at a time for two VLMs of one family (Qwen2.5-VL-7B, Qwen3-VL-8B). On RefCOCOg, what the selector reads matters more than how it is built: the VLM's image-conditioned query encodes the referent's location and beats a text-only query by $0.15$--$0.18$ with every head tested (by less for InternVL3-8B, whose vision encoder differs; one head tested). The remaining errors are rankings among a few top candidates; a cross-attention selector supervised with every valid box reaches the VLM's coordinate decoding on RefCOCOg, and with Qwen2.5-VL on RefCOCO/+ when also trained on their data. Beyond that distribution it does not: on Ref-L4's Objects365 images, mostly of categories outside training, every trained head falls to $0.19$--$0.28$ while Qwen decoding keeps about $0.8$ (InternVL3's drops too), neither a wider detector prompt nor a class-agnostic proposer closes the gap, and on public reasoning benchmarks Ref-Adv-s and RefBench-PRO full-budget heads trail decoding by $0.19$--$0.37$ although proposals cover most referents. A trained verifier reduces to an existence probe on the VLM's query state and does not beat the VLM's yes/no answer on public no-target benchmarks.

cs.CV

版本更新 · v2arXiv:2608.09482

Beyond Uniform Restoration: Empowering All-in-One Restoration with Pixel-Level Multimodal Guidance

Chunxiao Liu, Wei Liu, Anbin Xiong, Erli Meng

原文摘要

All-in-one image restoration is a unified low-level vision task that aims to effectively recover high-quality images from inputs degraded by various types and levels of corruption using a single model. Recent works have achieved remarkable progress by learning degradation-adaptive prompts or network architectures. However, these methods typically apply a uniform restoration strategy across the entire image, neglecting the fact that different regions may suffer from distinct degradation types and varying degrees of severity. In contrast, we propose to perform restoration at the pixel level, thereby enabling more fine-grained and precise control over the restoration process. Specifically, we present MGN-AIR, a novel pixel-level restoration framework for all-in-one image restoration. Our approach first learns to estimate a pixel-level visual prompt. Then, it leverages both textual and visual prompts to provide global and local degradation cues, guiding the model on where to look and how to restore at each pixel. We conduct extensive experiments on multiple all-in-one image restoration benchmarks, covering a wide range of tasks including denoising, deraining, deblurring, dehazing, desnowing, and low-light enhancement. Experimental results demonstrate that our proposed method consistently and significantly outperforms existing approaches.

cs.CV · cs.AI

版本更新 · v2arXiv:2604.07677

Bird-Inspired Spatial Flapping Wing Mechanism via Coupled Linkages with Single Actuator

Daniel Huczala, Sun-Pill Jung, Frank C. Park

原文摘要

Spatial single-loop mechanisms such as Bennett linkages offer a unique combination of one-degree-of-freedom actuation and nontrivial spatial trajectories, making them attractive for lightweight bio-inspired robotic design. However, existing synthesis methods rely on dual quaternion algebra and rational motion interpolation and factorization, placing task-based linkage design outside the reach of most engineers, a gap addressed by the Rational Linkages library recently developed by the authors. This paper presents a bird-inspired flapping-wing mechanism built from two coupled spatial four-bars, driven by a single motor. One linkage is actuated to generate the desired spatial sweeping stroke, while the serially coupled linkage remains unactuated and passively switches between extended and folded wing configurations over the stroke cycle. Additionally, we introduce a simplified kinematic methodology for constructing Bennett linkages from quadrilaterals that contain a desired surface area and further leverage mechanically induced passive state switching. This architecture realizes a coordinated sweep-and-fold wing motion with a single actuation input, reducing weight and control complexity. A 3D-printed prototype is assembled and tested, demonstrating the intended spatial stroke and passive folding behavior.

cs.RO

版本更新 · v2arXiv:2610.23908

ContraSafe-VL: A Contrastive Safety Preference Dataset for Vision-Language Models

Di Wu, Yanyan Zhao, Xin Lu, Hao Yang, Taotao Zhang, Yunzhen Zhao

原文摘要

Vision-Language Models (VLMs) are increasingly deployed as general-purpose multimodal assistants, making multimodal safety alignment an important challenge. Safety depends on the interplay between textual and visual content, with visual context often playing a critical role. However, VLMs may fail to use visual information appropriately, either overlooking safety-relevant visual content or over-interpreting sensitive visual content, leading to unsafe responses or over-refusal. Existing multimodal safety alignment datasets largely follow the response-preference paradigm, which teaches models how to respond but provides limited supervision on how visual context should inform the safety judgment, making it difficult to balance safety and over-refusal. To address this limitation, we introduce \textbf{ContraSafe-VL}, a \textbf{Contra}stive \textbf{Safe}ty preference dataset for \textbf{V}ision-\textbf{L}anguage Models. For each fixed textual query, ContraSafe-VL constructs paired visual contexts that result in multimodal queries with different safety labels, providing direct supervision for how visual context should inform multimodal safety judgments. It further provides response preference pairs based on harmlessness and helpfulness to encourage models to generate responses that are both safe and helpful. Experiments across three VLM architectures and multiple benchmarks demonstrate that training with ContraSafe-VL improves cross-modal safety recognition and achieves a better safety--over-refusal trade-off while preserving general multimodal capabilities. The dataset is available at https://github.com/pikepokenew/ContraSafe-VL

cs.CV

新论文 · v1arXiv:2611.11869

Action Representation for Spatial Generalization in Robot World Models

Yuhang Sun

原文摘要

Spatial generalization in robot world models depends on whether action representations capture the physical relationships that remain valid across changing spatial configurations. Yet different coordinates can describe the same interaction, whereas similar motions can produce different outcomes when local physical conditions change. How these distinctions affect learned prediction remains insufficiently understood. This study evaluates matched action representations defined in world, object, and surface coordinate frames using a controlled robotic peeling benchmark. Global pose and contact relocation experiments reveal a cost associated with limited training data when equivalent interactions receive different spatial descriptions. Removing information about peel thickness creates an irreducible ambiguity because identical predictive inputs can correspond to different material removal outcomes. Coordinates defined relative to the surface improve consequence prediction and counterfactual action selection under accurate local observations, but learned predictors remain vulnerable to uncertainty in the coordinate frame. Joint centering of the surface and tool removes sensitivity to translation of their common origin, whereas incorrect contact geometry and actual execution deviations introduce different failure mechanisms. These results show that effective spatial generalization depends on preserving the relationships between actions and states that determine physical consequences, rather than simply minimizing coordinate variation. This perspective provides a principled basis for designing robot world models that reuse physical knowledge across spatial configurations while retaining the distinctions essential to prediction and planning.

cs.RO

新论文 · v1arXiv:2611.11867

Competing Influence in LLM Populations: Budget Timing and Baseline Drift

Mehwish Nasim, Nicolas Fay

原文摘要

When competing campaigns target a population of large language model (LLM) agents, observed opinion changes may reflect both campaign influence and changes that would occur without intervention. We study how the timing of campaign expenditure and the population's own evolution shape outcomes in a network of persona-conditioned LLM agents that exchange and revise opinions in natural language. A proponent and a challenger are tasked with generating opposing messages under equal delivery budgets, while a separate stance probe measures opinions without influencing the simulation. Across eight claims, we compare single-campaign and competitive conditions against matched no-campaign runs, and evaluate all nine combinations of early, sustained, and late spending on a reference topic. Baseline stance drift varies substantially across claims and, on some, accounts for a sizeable fraction of apparent campaign effects. After accounting for this drift, the direction of observed campaign-associated changes aligns more closely with the claim's epistemic standing than with the campaigner's assigned role. Uninfluenced populations also consistently lose stance diversity, demonstrating that directional opinion change and convergence are distinct phenomena. In networks with two opinion-segregated communities, campaigns primarily reduce differences between communities rather than diversity within them. LLM campaign agents do not always argue for their assigned positions, particularly when these conflict with established evidence. Finally, campaign outcomes depend on the opponent's expenditure schedule: unlike a matched numerical bounded-confidence population receiving interventions at the same times and locations, the LLM population exhibits a payoff structure with no pure-strategy equilibrium.

physics.soc-ph · cs.CL · cs.MA

新论文 · v1arXiv:2611.11866

DeFSM: Robust and Efficient Decentralization of Federated Submodular Maximization

Duc A. Tran, Dung Truong, Duy Le

原文摘要

Submodular maximization seeks a small subset best representing a large dataset. In its federated setting, the server is a single point of failure and some clients may be Byzantine. Removing the server raises a trilemma of decentraliza- tion, efficiency, and security. We propose DeFSM, a Byzantine-robust and communication-efficient architecture that replaces the server with a lightweight smart contract and a pool of independent, consensus-free agents. Its aggregation rule, SafeGM, builds on two notions from combinatorial geometry: Tukey depth and safe points (the points lying in the convex hull of the good points wherever the bad points are). SafeGM makes the robust decision on 2D Gaussian random projections, which the contract can compute efficiently, and lifts it back as a convex combination of actual updates. Since lifting loses information, it uses the projected geometric median only when its depth certifies it, and otherwise a Tukey median. The share of bad updates is then bounded only by a count: how many there are, not what they contain. We prove, and confirm empirically, that this bound controls the solution quality. On CIFAR10 and PathMNIST, with up to 30% Byzantine clients and agents, DeFSM never collapses and retains on average about 91% of the quality of the ideal reference, the geometric median at a trusted central server. The geometric median decentralized in its best setting would retain about 96%, but the contract cannot compute it, and so each agent would have to download six times more data, or each client twice as much.

cs.DC · cs.LG

版本更新 · v2arXiv:2606.18671

HANSEL: Interactive Evidence for Post-Hoc Verification of Web Agents

Yujin Zhang, Yeonju Hwang, Daye Nam

原文摘要

AI web agents can perform complex tasks, such as searching for products and making purchases on users' behalf. However, verifying an agent's output remains difficult, because existing transparency mechanisms (e.g., full trajectory logs) require users to review logs and explanations, keeping verification a passive reading task. In this work, we view agent verification as an interactive activity and present HANSEL. Given an agent trajectory, HANSEL extracts evidence pages and snippets and presents them as navigable, live views that preserve relevant page state (e.g., applied filters), allowing users to verify how the agent arrived at its answer through interaction. In a controlled user study with 14 participants, we show that HANSEL significantly reduces task completion time and perceived effort, demonstrating that reframing verification as an interactive activity enables more efficient post-hoc verification of AI agents.

cs.HC

新论文 · v1arXiv:2611.11862

Are UFOs Real? Introducing Unified Factored Organizers

Thomas X. Sheppard, Alon Albalak, Joel Lehman, Kenneth O. Stanley

原文摘要

Implicit in much research in deep learning is the assumption that if a network achieves low loss, then it must have learned a good representation of its training data. Recent work by Kumar et al. (2025, arXiv:2505.11581) questions this assumption, arguing that networks trained with conventional gradient descent achieve low loss despite learning fractured entangled representations (FER), leaving related concepts disconnected and unrelated concepts entangled inappropriately. Kumar et al. further show that dramatically better representations are possible: a network evolved through an open-ended, human-in-the-loop process learned unified factored representations (UFR) that capture the regularities of the domain far better than a counterpart with nearly identical outputs trained with conventional gradient descent. Inspired by UFR, this paper introduces unified factored organizers (UFOs), networks trained to model regularities in the representations of other networks. A network can then be trained both to achieve low loss on its task objective and to align its representations with the model learned by the UFO. Networks trained with UFOs are shown to learn qualitatively superior representations in an image-regression domain where representations are easily interpretable. To provide quantitative evidence for the impact of UFOs on representation learning, a transformer is trained with and without UFO regularization to grok a modular arithmetic problem set. When combined with weight decay, UFOs can reduce the number of transformer training steps needed to grok by more than 10x relative to weight decay alone. Without weight decay, UFOs can reduce the number of transformer training steps needed to grok by more than 100x relative to an unregularized baseline. If UFOs are real, they suggest a new paradigm where loss functions themselves are trained to learn an effective model of good representation.

cs.LG