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 16:18 — 2026-11-16 16:18(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:2611.13560

Are Dense Rewards Worth Their Tokens in GRPO?

Vikash Singh, Weicong Chen, Debargha Ganguly, Yanyan Zhang, Mohsen Hariri, Sreehari Sankar, Wang Yang, Chaoda Song 等 19 位作者

原文摘要

RLVR trains language models to reason by rewarding correct final answers. A widely used algorithm, group-relative policy optimization (GRPO), gives every token of a solution the same credit, so dense process rewards have been proposed to credit intermediate steps, for example with a Monte-Carlo reward that scores each step by sampling continuations from it. Comparing such a reward with GRPO per training update, or vetting it with a gradient test on a single batch, leaves out what it costs: in our setting, the continuations alone take 91% of all generated tokens. We introduce a matched-compute protocol that gives every training intervention the same budget, measured in generated tokens and training GPU-hours, and tests it in preregistered studies with controls for side effects. Across 1,668 runs on six models, matching compute reverses the ranking. On Qwen2.5-1.5B, the Monte-Carlo reward leads standard GRPO per update, yet ends 14.3 percentage points lower at equal tokens and 7.8 points lower at equal training time, and none of four dense-reward variants outperforms GRPO even with as much training time. A per-token gradient test orders these rewards as training does but overestimates each by one to two points, as much as the gains it is meant to detect. In contrast, training on the prompts whose predicted success is nearest 1/2 improves accuracy at matched compute, by 1.7 points on Qwen3-0.6B and 0.8 on Qwen2.5-1.5B. These gains replicate on fresh seeds and unseen problems, and a simple rule, that the candidate prompts are predicted to fail more often than not, predicted where they appear in five of six preregistered tests. Both gains come early in training, and standard GRPO with twice the initial learning rate comes within half a point of them. We argue that training interventions should be validated at matched compute, which overturns some apparent gains and confirms others.

cs.CL · cs.AI

版本更新 · v2arXiv:2606.32012

CoMet: Context and Multiplicity Decomposition for Multimodal Uncertainty Estimation

Sanghyuk Chun, William Yang, Amaya Dharmasiri, Olga Russakovsky

原文摘要

Uncertainty estimation has been a long-standing challenge in AI models; it amounts to "knowing what you don't know," and metacognition is notoriously difficult even for humans (cf. the Dunning-Kruger effect). Although it is still far from solved even in simpler classification systems, tackling it in multimodal large language models (MLLMs) is becoming increasingly important. Within MLLMs, uncertainty can stem from any of the diverse sources as well as from their relationships, and further can stem from the unbounded answers in the open-ended setting. To tackle the issues, we propose CoMet, an MLLM uncertainty estimation method by decomposing uncertainty into a context-specific term and a multiplicity-specific term. The former captures ambiguity induced by the given context (e.g., task or prompt), while the latter captures how many plausible answers determined by the context remain compatible with the given input. We train a lightweight post-hoc uncertainty module to estimate these quantities, which enables efficient uncertainty estimation without autoregressive answer generation or repeated sampling. Experiments on various open-ended multimodal benchmarks, hallucination detection, and multiple-choice visual question answering benchmarks show that CoMet consistently improves uncertainty estimation over existing baselines while remaining efficient in practice. Code is available at https://github.com/princetonvisualai/comet_uncertainty

cs.LG · cs.CV

版本更新 · v2arXiv:2602.12253

Is Online Linear Optimization Sufficient for Strategic Robustness?

Yang Cai, Haipeng Luo, Chen-Yu Wei, Weiqiang Zheng

原文摘要

We consider bidding in repeated Bayesian first-price auctions. Bidding algorithms that achieve optimal regret have been extensively studied, but their strategic robustness to the seller's manipulation remains relatively underexplored. Bidding algorithms based on no-swap-regret algorithms achieve both desirable properties, but are suboptimal in terms of statistical and computational efficiency. In contrast, online gradient ascent is the only algorithm that achieves $O(\sqrt{TK})$ regret and strategic robustness [KSS24], where $T$ denotes the number of auctions and $K$ the number of bids. In this paper, we explore whether simple online linear optimization (OLO) algorithms suffice for bidding algorithms with both desirable properties. Our main result shows that sublinear linearized regret is sufficient for strategic robustness. Specifically, we construct simple black-box reductions that convert any OLO algorithm into a strategically robust no-regret bidding algorithm, in both known and unknown value distribution settings. For the known value distribution case, our reduction yields a bidding algorithm that achieves $O(\sqrt{T \log K})$ regret and strategic robustness (with exponential improvement on the $K$-dependence compared to [KSS24]). Moreover, we can further exploit the single-strategy benchmark and construct a bidding algorithm with $O(\sqrt{T(\log K+\log T)})$ regret and $O(1)$ strategic robustness. For the unknown value distribution case, our reduction gives a bidding algorithm with high-probability $O(\sqrt{T (\log K+\log(T/δ)})$ regret and strategic robustness, while removing the bounded density assumption made in [KSS24].

cs.GT · cs.LG

新论文 · v1arXiv:2611.13557

Darmok: Championship Cryptic Crossword Solving via Constraint Tightening and Resilient Orchestration

Siddhartha Chatterjee

原文摘要

We present Darmok, the first automated system to reach superhuman performance on British cryptic crosswords. Unlike American-style crosswords, cryptic clues disguise an unmarked definition and compositional wordplay beneath a misleading surface reading, sparse grids leave nearly half of each word's letters unchecked, and tournament rules rank a grid with even a single wrong or blank square behind all error-free solutions regardless of speed. Competing alongside an elite human field at the 55th Times Crossword Championship in London, Darmok solved all seven championship puzzles error-free and finished faster than every human competitor in every round. It solved the Final puzzle in 6m 52s, a full 3m 13s ahead of the human champion. Today's frontier AI models can on average solve a whole puzzle in a single try with comparable accuracy and speed. However, cloud API calls still stall, time out, trigger safety refusals, or get stuck in reasoning loops. To simultaneously achieve tournament-grade accuracy, speed, and reliability, Darmok cross-checks clues across three independent models in rounds; uses verified crossing letters to unlock harder clues; and uses a whole-puzzle model and dictionary lookups to fill any remaining gaps without overwriting confirmed words. This design maintained near-zero-error performance across a five-month daily evaluation period. Under normal benchmark conditions, removing backup models, fallback steps, or elaborate prompt instructions barely changes accuracy, making the architecture appear over-engineered. Yet during both the worst days of daily operation and a severe slowdown of our primary model in the live Championship Final, the backup paths carried the solve, showing that as foundation models master the average case, the central challenge in building complex AI systems shifts to resilience in the worst case.

cs.AI · cs.CL · cs.DC

新论文 · v1arXiv:2611.13555

Navigating relational agency in AI feedback musicianship: a diffractive dialogue

Yichen Wang, Charles Patrick Martin, David Gunawan, David Cooper

原文摘要

Musical AI systems behave in non-linear, dynamic ways that exceed the tool-oriented framings inherited from earlier instrument design. Existing frames of control and ensemble interaction do not articulate the entangled, relational agency that emerges when performers work with AI feedback instruments over time. We examine a duet practice with two AI-enabled feedback instruments, Sù and Agentier, in which feedback operates at the inner level of the AI's recurrent processes and the outer loop of musical interaction between performer, instrument, and co-performer. Our work draws on agential realism and a diffractive dialogue methodology to read our independent reflections through one another across five performances, examining how agency is enacted between performers and AI instruments. We find that control and agency are relational and shift across performances; musical textures enact feedback moments of control with performers; reaching a productive feedback environment is itself a shared exploratory act; and performers maintain their instruments through funnelling and injection, developing the skill of navigating agency along a continuous spectrum. Feedback musicianship with AI instruments demands navigation of this relational agency, displacing the leading-versus-responding binary that initially framed our reflections. This work contributes a materially-grounded vocabulary for AI feedback musicianship and a model of human-AI creative collaboration in which mutual agency is worked for through practice rather than designed in.

cs.SD · cs.HC

版本更新 · v2arXiv:2606.31682

HABIT: Human-Aware Behavior and Interaction Training Dataset for Robot Manipulation

Jaehwi Song, Suchae Jeong, Byeongguk Jeon, Sungdong Kim, Hyungmok Son, Kimin Lee

原文摘要

Large-scale demonstration datasets have been central to recent progress in general-purpose robot policies. However, existing datasets are collected in human-absent settings, and policies trained on such data may perform tasks competently in isolation but fail to exhibit human-aware behaviors. To address this gap, we introduce HABIT, a large-scale robot demonstration dataset for human-present environments. We organize tasks into three roles capturing distinct modes of human-robot interaction: Collaborator, where human and robot jointly accomplish a task; Coworker, where they pursue separate tasks in a shared space; and Supervisor, where the human directs the robot. The dataset comprises over 10K episodes and over 160 hours across 60 tasks. Our experiments show that training on human-present data elicits human-aware behaviors that robot-only data fails to produce: spatiotemporal synchronization in Collaborator tasks, yielding in Coworker tasks, and gesture grounding in Supervisor tasks. Moreover, training on HABIT enables rapid adaptation to new human-robot interaction tasks. By introducing human presence as a new axis of dataset diversity, HABIT extends robot policies to environments shared with humans.

cs.RO

版本更新 · v3arXiv:2602.02060

FiLoRA: Focus-and-Ignore LoRA for Controllable Feature Reliance

Hyunsuk Chung, Soyeon Caren Han, Seungyeon Ji, Jinwoo Kim, Eun-Jung Holden, Kyungreem Han

原文摘要

Multimodal foundation models integrate heterogeneous signals across modalities, yet it remains unclear whether their predictions can be controlled by explicitly modulating reliance on different internal feature pathways. Existing approaches to shortcut and spurious behavior primarily rely on post hoc analysis or data-level interventions, offering limited ability to directly intervene on how models use information. We introduce FiLoRA (Focus-and-Ignore LoRA), an instruction-conditioned, parameter-efficient adaptation framework that enables controllable modulation of feature reliance while keeping the task and predictive objective fixed. FiLoRA decomposes adaptation into feature-aligned low-rank modules and applies instruction-conditioned gating, allowing natural language instructions to act as computation-level control signals over internal representations. We evaluate FiLoRA across both controlled classification settings and generative multimodal tasks, and under a range of instruction types, including natural and compositional instructions. Results show that FiLoRA induces consistent and interpretable shifts in feature reliance, selectively amplifying or suppressing different feature groups in accordance with the instruction, without altering task semantics. Our findings suggest that instruction-conditioned parameter adaptation can serve as a practical mechanism for intervening on internal model behavior, providing a new perspective on controllability and analysis of multimodal systems beyond output-level prompting or post hoc interpretation.

cs.LG · cs.AI

版本更新 · v2arXiv:2611.11089

Control-Aware Command Generation for Asynchronous Control and Action Inference with Vision-Based Policies towards Fast Execution

Takayuki Osa, Kazuki Ota

原文摘要

Vision-based control policies, including vision-language-action models, have enabled robots to perform increasingly complex tasks. However, their execution speed is often limited by the high computational cost of action inference. In this work, we investigate what can be achieved at the control level without modifying the underlying policy learning process. We introduce an asynchronous inference and control framework that enables fast task execution while maintaining smooth and stable robot motions. By formulating the generation of continuous target positions from discontinuous model predictions as a tracking control problem, our method makes asynchronous inference practical for real robots. We further accelerate execution through adaptive action skipping based on the variance of predicted positions. Experiments with Action Chunking Transformer and OpenVLA-OFT on a bimanual robot demonstrate substantial speedups while preserving high task success rates.

cs.RO

版本更新 · v3arXiv:2603.01055

MMCOMET: A Large-Scale Multimodal Commonsense Knowledge Graph for Visual Reasoning

Jinwoo Kim, Rina Carines Cabral, Dhita Putri Pratama, Eileen Wang, Hiba Arnaout, Shuo Yang, Danyang Liu, Seungyeon Ji 等 13 位作者

原文摘要

Commonsense knowledge is widely used to support multimodal reasoning and generation, yet existing resources are predominantly text-only and lack grounding in visual context. This limitation often leads to the retrieval of semantically related but visually misaligned knowledge, and can degrade downstream performance. We introduce MMCOMET, a large-scale multimodal commonsense knowledge graph that integrates physical, social, and eventive knowledge with visual grounding. MMCOMET extends ATOMIC2020 by associating each commonsense triple with representative images through a hybrid alignment pipeline, resulting in over 900K multimodal tuples spanning 19 relation types. We further establish a unified evaluation framework across diverse vision-language tasks, including visual commonsense reasoning, visual question answering, visual storytelling, image editing, and image captioning. Our analysis shows that multimodal grounding provides more contextually aligned and reliable knowledge compared to both text-only knowledge graphs and LLM-based knowledge generation, particularly in reasoning-intensive scenarios. At the same time, we observe that the effectiveness of external knowledge is task-dependent, and naive knowledge integration can introduce noise in perception-dominated settings. To mitigate this, we propose an adaptive knowledge router that dynamically selects optimal paths or bypass knowledge in harmful settings. MMCOMET provides a scalable multimodal commonsense resource and a benchmark for studying the role of grounded knowledge in multimodal systems.

cs.AI

新论文 · v1arXiv:2611.13552

How Post-Training Influences Verbalized Algorithmic Computation in Language Models

Niall Dalton

原文摘要

Post-training improves language-model reasoning, but how it changes the use of internal information and verbalized chain of thought (CoT) remains unclear. We study successive SFT, DPO, and RLVR checkpoints of Olmo-3-7B-Instruct on multi-digit addition, using carry values as known intermediate state. Probes recover carry information from internal activations with similar accuracy in SFT and RLVR, although manipulating this information does not reliably control model outputs. In contrast, the usefulness of verbalized reasoning differs: CoTs produced by RLVR yield higher accuracy across receiving checkpoints than those produced by SFT or DPO. We also show that carry-related statements receive higher importance scores after RLVR than matched random controls. DPO shows similar sensitivity to carry-related statements but without comparable accuracy gains. These results suggest that post-training changes the usefulness of written reasoning and sensitivity to its content in meaningful ways, although greater sensitivity does not necessarily coincide with better problem performance.

cs.CL · cs.LG

版本更新 · v2arXiv:2611.10168

Key Institutional Considerations for Designing Effective Independent Verification Organizations

Uma Kalkar, Rumman Chowdhury, Reva Schwartz, Gabriella Waters

原文摘要

Independent verification organizations (IVOs) have emerged as a popular, bipartisan solution for facilitating high-quality independent evaluations of AI technologies. But momentum is outpacing design: there is emerging consensus on the necessary operating conditions for robust IVOs, but little discourse on institutional and policy specifications to ensure those conditions are met. IVOs are currently envisioned primarily as a mechanism operating at the AI frontier and model level to evaluate catastrophic risks. Little work has yet been done on how to establish or enforce an IVO ecosystem that is equipped to handle myriad risks (i.e., beyond catastrophic ones) and operate at the system and application level across the deployment and use stages of the AI lifecycle. This paper argues that IVO design should ensure the capacity for evaluating AI models, systems, and applications in context for both frontier companies and enterprise companies & AI-adopting organizations. Additionally, this paper provides early considerations to advance IVO discourse from normative to operative, and from existential to everyday risks. Drawing from existing institutional constructs in financial services, education, public health, and science, we outline key considerations on whom IVOs serve, what access they require, how they interact with external governing structures, how they are funded, and how they fit into the wider assurance ecosystem. Much of the structure and knowledge necessary to holistically design IVOs and their ecosystem does not yet exist, even as companies rush to ink agreements with their chosen evaluators.

cs.AI

版本更新 · v3arXiv:2607.16826

Spatially-Aware Class-Agnostic Object Counting

Robert Wijaya, Md. Tanvir Hossain, Amanda Kau, Ngai-Man Cheung

原文摘要

Generalised object counting aims to estimate the number of instances of an arbitrary object category from a single image, but many recent methods can struggle on structurally complex objects due to limited spatial modelling. We present \textit{UpCount}, a class-agnostic counter designed to better preserve spatial structure. UpCount strengthens the visual representation by extracting multi-layer features from a ViT-B/16 encoder and reassembling them into a refined multi-scale pyramid that is spatially refined using Dense Prediction Transformers and FeatUp, yielding features with improved structural and spatial sensitivity; a proposal--verification counting head then identifies repeated patterns and produces a density map for the final count. On FSC-147, UpCount achieves 12.39 MAE and 100.89 RMSE on the test set, and it transfers effectively to vehicle counting on CARPK (6.27 MAE, 8.79 RMSE). Code: https://github.com/r28112072-rgb/upcount

cs.CV

版本更新 · v4arXiv:2606.24548

Are Text-to-Image Models Inductivist Turkeys? A Counterfactual Benchmark for Causal Reasoning

Jiayi Lei, Yuandong Pu, Xingyu Han, Rongpeng Zhu, Jing Xu, Jinyao Wang, Zijian Zhou, Bin Fu 等 11 位作者

原文摘要

Text-to-image (T2I) generation models have achieved remarkable progress in producing visually realistic images from natural language prompts. Yet it remains unclear whether their success reflects genuine causal understanding or sophisticated pattern matching over visual-textual correlations. Inspired by Russell's inductivist turkey, we introduce Counterfactual-World (CF-World), a counterfactual benchmark designed to investigate whether text-to-image models can generate images under rules that systematically contradict real-world priors. CF-World organizes each scenario into three progressive levels: factual generation under ordinary world knowledge, explicit counterfactual generation with direct visual instructions, and implicit counterfactual generation requiring causal deduction from altered rules. We evaluate open-source and closed-source text-to-image models using CF-Eval, a scoring pipeline based on vision-language models. We report the scores at each prompt level as the primary results and use Prior Resistance Rate (PRR) and Reasoning Retention Rate (RRR) as auxiliary descriptive indicators. All evaluated models score lower on counterfactual prompts than on factual prompts. Diagnostic experiments reveal difficulties in following altered rules and recombining rarely co-occurring visual attributes. The gains from rephrasing object and attribute names are consistent with a role for learned visual priors in counterfactual generation. Our benchmark and evaluation code are publicly available at https://jylei16.github.io/CF-World.github.io/.

cs.CV

新论文 · v1arXiv:2611.13550

Tabular Foundation Models in Financial Predictions

Ed deHaan, Allen H. Huang

原文摘要

This technical note introduces tabular foundation models (TFMs) to accounting research and evaluates their performance in one common benchmark task: predicting firms' next-year earnings changes. TFMs produce numeric predictions and learn in context, without post-release training or tuning, making them natural and easy to use for forecasting. The TFMs we study are pretrained on synthetic tabular data, so they avoid look-ahead bias from memorizing firms' outcomes. We compare Google Research's TabFM with a fine-tuned open-weight large language model developed in recent accounting research and with several tree-based models. TabFM is more accurate than all comparison models. It is also faster and less computationally intensive than the large language model. Several tree-based specifications produce predictions more quickly than TabFM, so choosing between them involves a trade-off between accuracy and speed. We conclude that TFMs have promise as a new tool for forecasting in accounting and finance. However, our evaluation is based on a single benchmark task, and no class of models is likely to dominate in all settings.

cs.LG · q-fin.CP · q-fin.ST

新论文 · v1arXiv:2611.13549

Perspicacious: An Educational Intelligence Platform for Instructional Designers

Yumou Wei, Michael C Sutton, Paulo F. Carvalho

原文摘要

Designing effective instruction requires instructional designers to create assessment questions aligned with clearly defined knowledge components (KCs). These KCs represent specific knowledge or skills that students must acquire. This process is often time-consuming and relies heavily on expert labor. Although AI for education research has produced a growing body of automated question-generation and KC-discovery methods, few have moved beyond research prototypes into user-friendly tools that instructional designers can readily use. To bridge the research-to-practice gap, we developed and evaluated an educational intelligence platform that unifies question generation and KC discovery behind a single interface to support instructional content development. Named "Perspicacious," our platform aims to help instructional designers "see through" complex data and algorithms to derive instructional insights. A cost model calibrated on 17 production runs revealed a cost-time trade-off: extracting KCs from a 576-question bank cost US $1.98-4.11 and took 36-158 minutes to complete, depending on the budget. A formative, partially Wizard-of-Oz study with eight instructional designers found strong perceived usability (mean SUS 78.8) and surfaced concrete requirements that we treat as prerequisites for deployment.

cs.CY · cs.ET · cs.HC

新论文 · v1arXiv:2611.13548

Massive Activations Confound Sparse Autoencoders

Nathan Delisle, Chenhao Tan

原文摘要

Sparse autoencoders (SAEs) are commonly evaluated by variance explained and ablation damage. We show how massive activations undermine both evaluations, and propose a correction. Massive activations cause up to 99.9% of activation variance to lie along the first principal component (PC1). Because the PC1 component has no unique sparse decomposition, each SAE run rebuilds it from a different set of features. These features dominate importance rankings: an average of 47 of the 50 most important features are unmatched in a separate run, compared with about 27 of a random 50. A dictionary can also report 99.96% of variance explained while leaving 33% of the variance outside PC1 unexplained. We bypass PC1 by training the SAE on activations with PC1 removed, then adding PC1 to the reconstruction. In affected models, the top-50 feature sharing rate rises from 4-39.5% to 73-99.5%. At the same TopK sparsity, the resulting SAE degrades the model up to 44% less and reduces content reconstruction error by up to 33.9%.

cs.LG · cs.AI

版本更新 · v3arXiv:2610.27745

Hearsay or Fact? How Framing and Elicitation Shape LLM Probability Updates

An-Sheng Jhang

原文摘要

We measure LLM probability updates on $δ$-SNLI premise-hypothesis pairs across the Multi-Candidate Verbalization (MCV) family. Separate calls elicit priors and posteriors; directional movement s signs changes by the annotated direction. We cross stated vs. reported wording with Direct single estimates (D0) vs. Multi-Candidate Weighted elicitation (MCW; five weighted estimates, V0, weighted mean M_W). Frozen formal tests cover 15 configurations (23,400 planned units): 3 primary (Qwen3-8B thinking off, OLMo-2-7B, gpt-6.1-sol via codex; 6,600), 4 CLI-harnessed configurations (gpt-oss-120b, claude-sonnet-5.5, gemini-3.8-flash low/high; dnli-agy; 5,600), and 8 un-harnessed REST API configurations (dnli-api; 11,200). We add a 20-configuration, 8-provider dev ladder (3B-550B; 6,000 units) and a 2-model OpenAI 1+1+4 ablation of D0, Direct-CoT (D_CoT), MCW (V0), and Multi-Candidate Unweighted (MCU; V2 / M_U; 1,800 calls). All 15 reject H1 (s = +16.1 to +28.9 points) and H4a (endpoint reduction +0.065 to +0.391); 13 reject H2 (+2.0 to +8.9 points), except gpt-oss-120b (+0.9) and sf:qwen3.5-4b (-0.64). In 14 of 15 formal configurations, readouts obey Top-1 > MCW (M_W) > MCU (M_U). On 120 common pairs, API Qwen3 tracks local Qwen3 (H1 +18.72 vs. +18.90; H4a +0.265 vs. +0.254; H2 +3.78 vs. +1.67; H4b -9.03 vs. -6.70); local H2 rejects on all 240 pairs (+1.97), not this subset. API Gemma 31B (formal H4b -2.85, Top-1 +0.64; spread -13.3% / +7.6%) and Groq gpt-oss-120b (dev -4.15, +0.40; -18.4% / +12.2%) show small MCW costs and stated narrowing/reported widening without CLI harnesses. In OpenAI ablations, D_CoT keeps numeric-token entropy <1 bit; MCU (V2) shows smaller movement loss but halves spread and $ΔH_2$ vs. MCW (V0). Because $δ$-NLI annotates only direction, we make no claim about appropriate update magnitude. Code/data: https://github.com/asj252/llm-judgment-studies

cs.CL · cs.AI

新论文 · v1arXiv:2611.13546

Statistical Analysis of the Structure from Noise Phenomenon in Generalized Procrustes Problems

Jiayang Yin, Shuyang Ling

原文摘要

In cryo-electron microscopy (cryo-EM), observations are highly noisy, and scientists often use reference templates to guide matching, alignment, and averaging to reconstruct the molecular structure. However, these procedures can produce an output resembling the template even when the observations contain only noise, a phenomenon known as Einstein from noise (EfN). This phenomenon has attracted attention in statistics, where simplified observation models involving group actions have been studied to understand the resulting model bias. Characterizing the bias in the full cryo-EM pipeline remains challenging. We study template matching and averaging in the generalized orthogonal Procrustes problem, which provides an analytically tractable setting involving continuous orthogonal group action and lower-dimensional projections of ordered point clouds. We show that, for pure Gaussian noise, the mean of the aligned-and-averaged estimator admits the same left and right singular-vector bases as the template but its singular values may be distorted. This identifies the subspace structure imposed by the template even in the absence of signal. We characterize the singular values and bound their distortion, establish finite-sample operator-norm bounds, and give sufficient sample sizes for the population mean to dominate random fluctuations. In fact, this bias still exists under weak signals: when the template rank is at least the observation dimension, the change from the pure-noise mean is $O(η^2)$ in signal amplitude $η$. Our results provide an explicit characterization of model bias in a registration problem motivated by cryo-EM and show why averaging more observations can stabilize structure imposed by the template without removing the bias.

math.ST · cs.IT

新论文 · v1arXiv:2611.13545

Closing the BF16-FP16 Convergence Speed Gap in RL Post-Training with Stochastic Rounding

Mahdi Nazemi, Oleg Rybakov, Tijmen Blankevoort, Asit Mishra, Carlo del Mundo

原文摘要

BF16 is widely used to pretrain large language models because it has the same dynamic range as FP32 and converges similarly to FP32 training. However, in reinforcement learning with verifiable rewards, where learning rates are typically low, we find that standard mixed-precision BF16 training converges more slowly than FP16 training at the same learning rate. We attribute this gap to weight sticking: when FP32 shadow weights are quantized to BF16 using round-to-nearest, small updates often leave the BF16 weights used for rollout generation and training unchanged. Consequently, several optimizer steps may be needed before updates to the shadow weights are reflected in the BF16 weights. Raising the BF16 learning rate reduces weight sticking and matches FP16's convergence speed, but it also increases the KL divergence from the reference model. We instead apply stochastic rounding when quantizing the FP32 shadow weights to BF16. This approach matches FP16's convergence speed across different models and datasets while reducing the KL divergence relative to BF16 training with a larger learning rate. Unlike FP16, it keeps BF16's wider dynamic range and needs no loss scaling.

cs.LG · stat.ML

新论文 · v1arXiv:2611.13544

DASWeave: Weaving Evidence across Speaking Conditions for Multimodal Depression, Anxiety, and Stress Screening

Binyan Chen, Tong Pan, Yadong Liu, Xin Yun, Shangfei Wang

原文摘要

Multimodal screening for depression, anxiety, and stress in children and adolescents uses verbal and nonverbal behaviour. Most existing multimodal methods learn from a single recording or interview. Yet a participant's behaviour changes with the speaking condition, and these changes hold evidence that is invisible in a single recording. We propose DASWeave to bring this evidence into a neural network that encodes each response and then combines all responses of a participant. We train supervised trees to find which behavioural statistics matter for screening, both within each response and in how they change across conditions. We feed the tree outputs into the network both before and after this combination. We fine-tune a large language model to relate what a participant says to how they say it across conditions, and feed its representation into the network after the combination. The language model reads behaviour as text descriptions. We therefore derive additional training targets for the network from a multimodal large language model that takes in the audio-visual features directly. On the official hidden test set of the ACM Multimedia 2026 AdoDAS Grand Challenge (Track 1), DASWeave ranks second among 95 eligible teams. Its mean F1 is 27.1% higher than that of the official audio-visual baseline. It also has the highest mean AUROC among the top five teams. Ablation studies on the validation set show that the supervised trees and the fine-tuned language model contribute most to mean F1.

cs.MM

新论文 · v1arXiv:2611.13543

Sanity Checks for Attention Surgery: Null Models and Matched Interventions for Representation-Motivated Design

Basavaraj R M, Aryan Bharat Kumar

原文摘要

Transformer architectural modifications frequently measure a statistic on internal representations, observe that it is large, remove it architecturally, and report an improvement. Interpreting that sequence requires evaluating whether the statistic exceeds its noise floor, whether it survives a geometric null, and whether the training gain requires the specific direction removed. We formulate these requirements as three reusable checks and apply them to self-attention removal, which deletes the component of an attention head's output aligned with its own value vector. Across nine multi-head pretrained models (5,408 heads, 124M to 6.9B parameters), an earlier-position causal null accounts for 48% to 63% of the head-value cosine alignment, with no monotone scale trend. Under grouped-query attention, this alignment is specific to each head's own key-value group and reverses across groups (+0.24 to +0.27 within-group, -0.11 to -0.19 across-group). Split-half reliability demonstrates that the per-head intervention effect is resolvable in GPT-2 ($r_Δ= 0.80$) but attenuated in Pythia-160M (0.42) and Pythia-410M (0.53), whereas structural self-cosine reliability is near-perfect in all three ($r_{\text{stat}} > 0.99$); which statistic predicts its own intervention is model-dependent. In a pre-registered 3-arm by 8-seed paired factorial at $4.0 \times 10^8$ tokens per run (399,900,672 tokens), a fixed arbitrary rank-one direction does not differ from baseline (+0.0011 nats, $p = 0.197$), while self-directed removal lowers validation loss (-0.0029 nats, Holm-corrected $p = 0.0023$) and outperforms the fixed arbitrary control directly (-0.0040 nats, $p = 0.0018$). At 50.9M scale the training benefit requires the own-value direction rather than a fixed arbitrary one, but a large cosine alignment alone does not establish redundancy in frozen models.

cs.LG · cs.CL

版本更新 · v2arXiv:2611.04748

Mass Scaling Laws for Energy-Optimal Horizontal Multirotor Flight

Ayush Patnaik, Nicolas Michel, Leonard Baursfeld, Davide Scaramuzza, Xinfan Lin

原文摘要

Multicopters are a type of aerial robot seeing extensive real-world applications but with range and endurance as critical bottlenecks. Optimization of its energy efficiency through energy-aware flight planning (e.g. cruise speed-scheduling) has hence been an important topic. The mass of the multicopter is an important factor to consider, as it has a significant impact on efficiency, and many common applications are subject to frequent mass variation due to payload change. This work derives closed-form scaling relationships between the energy-optimal steady horizontal flight behavior of a multicopter and its vehicle mass, and validates them experimentally on two multirotor platforms. Specifically, for mass-only variation, i.e., when payload changes the vehicle mass while the airframe geometry, aerodynamic characteristics, and rotor configuration remain unchanged, we show through a combination of numerical simulation, theoretical derivation, and experimental validation that (1) the energy-optimal velocity is proportional to the square root of vehicle mass; (2) the energy-optimal pitch angle is constant with respect to mass; and (3) the optimal energy per meter is proportional to mass. The analysis shows that these three relationships arise from a common mass-similarity property of the steady-state flight and rotor-aerodynamic equations. Furthermore, the obtained scaling relationships enable simple yet effective cruise-speed planning to optimize energy efficiency and extend the range and endurance of the multicopter.

cs.RO · eess.SY

版本更新 · v2arXiv:2611.11399

RoCoMark: Robust Code Watermarking Against LLM-Based Functionality-Preserving Rewriting Attacks

Hongjie Zhang, Xingcheng Li, Hanzhou Wu, Zhihua Xia, Xinpeng Zhang

原文摘要

Large Language Models (LLMs) present significant challenges to source tracing of generated code, as LLM rewriting can substantially alter the syntactic and structural characteristics of watermarked code. However, existing watermarking methods often suffer severe performance degradation under LLM-based functionality-preserving rewriting attacks. We attribute this vulnerability not merely to the intensity of rewriting, but to a fundamental distribution mismatch between the perturbations encountered during watermark training and the global, non-deterministic rewriting behaviors induced by LLMs. In order to deal with this challenge, we propose RoCoMark, a Robust Code waterMarking framework that explicitly learns and adapts to LLM rewriting behaviors. RoCoMark is a multi-round minimax game between the watermark embedder and LLM rewriter. In each round, the embedder first learns threat profiling from LLM-rewritten code by optimizing the watermark encoder through contrastive learning and knowledge distillation while mitigating catastrophic representation drift. Then the embedder improves strategies by optimizing the watermark encoder and extractor simultaneously. Finally, an adaptive attack-evolution mechanism based on Reflexion and short- and long-term memory pools is applied, enabling continuous interaction between the watermark embedder and evolving rewriting strategies. This iterative attacker-defender process progressively improves the robustness of the watermark against diverse and previously unseen LLM rewriting behaviors. Extensive experiments on four typical benchmarks, under attacks generated by five LLMs, show the effectiveness of RoCoMark. RoCoMark achieves an average detection AUROC of 74.2% and a Rewrite BitACC of 75.3%, while maintaining a code functionality pass rate above 99.7%.

cs.CR

新论文 · v1arXiv:2611.13541

Agent Benchmarks Stop Ranking Before They Run Out of Tasks

Zihan Dong, Qishi Zhan, Minxuan Hu

原文摘要

Data-science and ML-engineering agents are evaluated on benchmarks such as MLE-bench, ALE-Bench, SWE-fficiency and AlgoTune, whose tasks yield a quantity (a Kaggle metric, contest score or speedup). When one stops separating the best models, the usual remedy is a harder version. Re-scoring released runs of eight benchmarks, including SWE-bench Verified and Terminal-Bench for breadth, we find much of this saturation lies in the measurement, not the tasks. Most agent benchmarks reduce a run to one pass/fail bit, which fails in two ways. It is too coarse: neighbouring top-ten agents on SWE-bench Verified differ by less than its tests' errors, and correcting them reorders neighbours no more than deleting a random sixth of each agent's passes. It is capped at the reference: ALE-Bench's top ten, separated by contest score, tie when graded against the median human, although nearly half the runs beat that human. On fixed tasks, a higher bar only moves the problem; graded credit fixes it. We derive a clipped log-ratio score with dated anchors and cost as a separate axis. In a Terminal-Bench re-run, pass/fail-tied models differ over tenfold on the quantity the tests discard.

cs.AI

新论文 · v1arXiv:2611.13540

AgenticConflictLens: From Long Traffic Videos to Verifiable Conflict Evidence

Boshra Khalili, Sonia Moshfeghi, Jinwoo Jang, Andrew W Smyth

原文摘要

Traffic cameras continuously record urban road-user behavior, providing proactive safety evidence beyond crash reports. Yet safety-critical events - conflicts, near misses, unsafe turns, sudden braking, and crosswalk interactions - are rare and brief, buried in hours of routine footage. Existing traffic-video pipelines extract detections, trajectories, and risk scores but process footage uniformly, without adapting to an analyst's query. Vision-language methods improve scene understanding but often lack traffic-specific verification of participant-level conflicts. We introduce AgenticConflictLens, an agentic, query-conditioned framework for efficient, auditable conflict extraction from long fixed-camera traffic videos. Given a natural-language safety query, a Traffic-Aware Controller combines structured traffic memory with specialized conflict-analysis tools to choose where to search, which road users to examine, and which evidence to verify or reject. A coarse-to-fine search retrieves candidate intervals and applies detailed verification selectively. The framework produces conflict records with refined time intervals, participant pairs, semantic regions, trajectories, motion evidence, uncertainty, rejection reasons, and supporting clips. On 96 annotated hours from Bellevue and NYC DOT, AgenticConflictLens achieves the highest conflict-existence F1 and temporal IoU on both datasets: 0.84 F1/0.77 tIoU on Bellevue and 0.80 F1/0.73 tIoU on NYC DOT, while inspecting only 8.7% and 11.9% of video, respectively. These results support scalable, evidence-grounded urban traffic-safety analysis.

cs.CV

新论文 · v1arXiv:2611.13539

Exploiting Feature-invariant Scaling Laws for Efficient A/B Experimentation in Recommendation Systems

Zhiliang Chen, Sammy Nammari, Jiahao Su, Xiaoqiang Luo

原文摘要

Researchers and engineers improve recommendation systems by performing A/B experiments with new model architectures, optimization methods, and data features. However, running full-scale experiments (e.g., full-model training with online data) for every feature change is prohibitively computationally intensive, while the result from a single small-scale experiment is often misleading. To address this challenge, we show that modern recommendation models exhibit feature-invariant scaling: the change of a model's performance follows a predictable scaling-law even when different feature changes are applied to the baseline model. We introduce an efficient method that exploits scaling-laws and uses statistical inference to predict (with confidence intervals) the full-scale effect of an A/B experiment in recommendation systems. This enables practitioners to efficiently filter out neutral or regressive feature changes with only small-scale experimentation. We validate our method on three distinct recommendation model families---Wukong, HSTU, and LLM-based recommendation models---across more than 40 architectural, optimization, and data changes. Our method accurately distinguishes beneficial feature changes from neutral and regressive ones while using only a fraction of the compute required for full-scale evaluation. It also identifies several changes that improve HSTU's state-of-the-art performance on MovieLens-20M. Our results demonstrate that scaling laws can be practically exploited to make compute-efficient decisions for improving industrial-scale recommendation systems.

cs.IR · cs.LG

版本更新 · v5arXiv:2512.20113

Attention Fusion for Bridge Deck Delamination Detection

Alireza Moayedikia, Amirhossein Moayedikia

原文摘要

Subsurface delaminations in reinforced concrete bridge decks escape conventional visual inspection, and the two principal sensing techniques used to find them are individually incomplete: Ground Penetrating Radar (GPR) penetrates deeply but degrades near the surface, while Infrared Thermography (IRT) resolves shallow defects but cannot reach deeper structure. This paper presents a framework for fusing the two modalities through hierarchical attention: temporal self-attention over GPR A-scans, channel-spatial attention over IRT patches, and cross-modal multi-head attention with learnable modality embeddings, coupled with decomposed aleatoric/epistemic uncertainty estimation. Beyond the architecture itself, which is lightweight at approximately 0.53M parameters with a closed-form accounting of where capacity resides, we contribute an elementary formal analysis. Two-token cross-modal attention is shown to be exactly a bank of per-sample learned gates; a gradient-allocation proposition quantifies how class imbalance starves attention parameters of minority-class signal and how loss reweighting trades that starvation for gradient variance; and closed-form metric floors under majority-class collapse anchor a diagnostic divergence between ranking metrics (AUC) and thresholded metrics (F1). The analysis suggests that adaptively weighted fusion, precisely because its feature-selection policy is learned, may be distinctively vulnerable to the severe class imbalance typical of operational bridge decks; establishing whether and when this occurs is deferred to empirical evaluation.

cs.CV · eess.IV

新论文 · v1arXiv:2611.13536

Adaptive catalyst discovery emerges from composable scientific intelligence

Wentao Li, Qilong Cai, Qi Lei, Jinxing Chen, Ruixuan Chen, Jiangjie Qiu, Botian Wang, Kai Zhao 等 24 位作者

原文摘要

Catalysis underpins chemical manufacturing, yet knowledge for catalyst design remains dispersed across experimental observations, computational data and catalysis subdisciplines. To overcome this challenge, we present Composable Catalytic Intelligence (CatAI), a physics-grounded modular AI framework that learns 1.8 million structure-property relations and organizes them into 30 physicochemical encoders. This reusable intermediate layer enables task-specific models to guide catalyst discovery for targeted reactions and objectives. Across benchmark tasks, CatAI achieves a 7.8% lower task-averaged prediction error than the next-best models. In experimental validation, CatAI identified a new confined-passable pocket catalyst applicable in industrial homogeneous polymer synthesis with 9 times the activity of the commercial benchmark. In parallel, it guided autonomous closed-loop optimization of a highly active Cu-based geminal-atom photocatalyst for C-O coupling reactions surpassing the traditional Ni-based systems. This approach connects property-level data to catalytic objectives, supporting physicochemical knowledge reuse across mechanistically distinct systems through computational screening and iterative model-experiment feedback.

cs.LG

新论文 · v1arXiv:2611.13534

Beyond the Log-Linear Frontier: Pre-Asymptotic Analysis of Finite-Budget Neural Scaling Laws

Tianyu Zhao, Zhixiang Ren

原文摘要

Classical neural scaling laws model excess risk as a separable power law in model capacity $N$ and dataset size $D$. At finite budgets, however, empirical learning curves and compute-optimal allocations depart from this log-linear frontier. This paper analyzes how spectral and domain heterogeneity govern these finite-scale departures. Along the capacity axis, Mellin--Barnes analysis of covariance resolvents yields a three-term expansion where competing spectral and regularization modes produce finite-scale curvature and logarithmic transitions. Because fitting multiple power laws on narrow pilot grids is ill-conditioned, we eliminate the loss floor via first-order Aitken differencing to obtain a well-conditioned closed-form initializer. Along the data axis, KKT duality reduces $K$-domain mixing to scalar water-filling, yielding finite-budget bottleneck bounds, an online sampling curriculum, and a simplex Hessian metric for cross-domain transfer. Along the iso-compute frontier $C=6ND$, we prove that the optimal allocation exponent drifts in proportion to capacity and data exponent variances, while Schur-complement reduction separates intrinsic drift from frozen-hyperparameter bias. Across 45 operator spectra, Pythia (70M--12B), Chinchilla (245 runs), and RegMix (17 domains), our estimators achieve $5.89\%$ holdout MAPE, cut high-compute allocation regret by up to $76.4\%$, and reduce cross-scale mixture RMSE by $59.4\%$ with zero selection regret.

cs.LG

新论文 · v1arXiv:2611.13531

Framing-Induced Vulnerability in Large Language Models: Evidence from Chinese News Headlines

Mo Zhou, Yiwei Zhang

原文摘要

Large language models increasingly evaluate, route, and moderate news, so a model's judgment of a headline has become a gatekeeping decision. We ask whether that judgment follows the fact being reported or the way the fact is framed. We combine an ecological audit of 63,926 Chinese news headlines with a controlled experiment. The experiment covers 30 real events, 30 synthetic events, and 29 natural headline pairs, tested on four Chinese models. Changing only the headline frame shifted both perceived misleadingness and event severity, and the direction and ordering of the shifts were the same for DeepSeek, MiMo, Dots, and LongCat. Emotionally intensified headlines raised severity ratings by 0.65 to 0.88 points, whereas positively reframed headlines lowered them by 0.34 to 0.52 points. Supplying the article body reduced this severity bias by 54 to 80 percent, and all four reductions were significant at p < .01. A small human reference study showed the same direction of effect. Headline-only LLM evaluation therefore carries a measurable robustness gap, and framing sensitivity belongs among the reliability properties that deployment and governance check.

cs.CL