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

+1680新增论文
+1333其中 AI 相关
+187已有论文修订
+59新增 AI 导读

arXiv 最近完成扫描:2026-11-18 04:28 · 全站最近入库:2026-11-17 23:35各信源进度 →

统计口径与数据新鲜度

统计 2026-11-17 09:05 — 2026-11-18 09:05(北京时间) · 点击数字查看对应简报;AI 相关包含在新增论文内。

最近完成检查:2026-11-18 04:28最近发现新增版本:2026-11-17 23:35库内论文最新更新时间:2026-11-17 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。

25,786 已收录论文21,491 AI 相关7,725 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 5,970 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-18 04:27(北京时间)
版本更新 · v4arXiv:2604.05164

Not All Turns Are Equally Hard: Adaptive Thinking Budgets For Efficient Multi-Turn Reasoning in Agents

Neharika Jali, Anupam Nayak, Gauri Joshi

原文摘要

As LLM reasoning performance plateaus, improving inference-time compute efficiency is crucial to mitigate overthinking and long thinking traces even for simple queries. Prior approaches including length regularization, adaptive routing and difficulty-based budget allocation primarily focus on single-turn settings and fail to address the sequential dependencies inherent in multi-turn reasoning. We formulate multi-turn reasoning as a sequential compute allocation problem and model it as a multi-objective Markov Decision Process. We propose TAB: Turn-Adaptive Budgets, a budget allocation policy that takes as input the conversation history and learns to maximize task accuracy while respecting global per-problem token constraints, and consequently learns to adaptively allot smaller budgets to easier turns and save an appropriate number of tokens for the crucial harder reasoning steps. We train and analyze the behavior of TAB on an auxiliary hierarchical plan-and-execute system for multi-turn math reasoning where solution trajectories are inexpensive to collect. Here, TAB saves up to 35% tokens and 30% latency over static and off-the-shelf LLM budget baselines. We then transfer the same TAB policy to environment-grounded agentic workflows spanning sequential tool-use, simulated user-agent dialogue and multi-step terminal tasks, where it generalizes and maintains accuracy while saving up to 20% tokens and 15% latency, without ever being trained on them.

cs.LG · cs.AI

版本更新 · v3arXiv:2601.03708

MHRC-Bench: A Multilingual Hardware Repository-Level Code Completion benchmark

Qingyun Zou, Jiahao Cui, Nuo Chen, Bingsheng He, Weng-Fai Wong

原文摘要

Large language models (LLMs) have achieved strong performance on code completion in general-purpose programming languages, but existing repository-level benchmarks focus almost exclusively on software code and largely overlook hardware description languages (HDLs). We present \textbf{MHRC-Bench}, consisting of \textbf{MHRC-Bench-Train} and \textbf{MHRC-Bench-Eval}, the first benchmark for multilingual hardware code completion at the repository level with functional evaluation. MHRC-Bench covers three major hardware design coding styles and annotates each completion target with code-structure-level and hardware-oriented semantic labels derived from concrete syntax tree (CST) analysis. Post-training billion-scale code LLMs on MHRC-Bench-Train enables smaller open-weight models to outperform much larger general-purpose models under matched context windows. Beyond aggregate accuracy, our analysis surfaces four findings that distinguish hardware from software code completion: \emph{(i)}~performance varies sharply across HDLs, with high-level synthesis (HLS) easiest and VHDL hardest in most settings across functional pass rate, exact match (EM), and edit similarity (ES); \emph{(ii)}~hardware accuracy is non-monotonic in code-structure depth, in contrast with the monotonic depth--accuracy curve known for software; \emph{(iii)}~accuracy plateaus once the input context exceeds 2{,}048 tokens, while software completion continues to benefit from longer context; and \emph{(iv)}~certain repository-level retrievers tuned on software underperform a no-retrieval baseline on HDL repositories because their structural and dependency assumptions are violated by hardware code.

cs.PL · cs.AI

版本更新 · v3arXiv:2610.08586

MINDSET: Energy-based Memory Evolution for Long Conversational Agents

Sujato Dutta, Sreekruthy Tummala, Shashank Vanga, Ayushmi Pavani, BHVSP Subrahmanyam, Neeraj Choudhary

原文摘要

Long-term conversational memory requires more than retrieving past dialogue. A system must determine whether new evidence replaces an existing fact, applies in a different context or represents an unresolved contradiction. We introduce MINDSET, a local memory controller that makes these decisions without any LLM calls. It preserves original dialogue episodes and organizes them into provenance-linked, versioned schemas. For each incoming episode, a feasibility-constrained energy objective governs four possible transitions: reinforcement, supersession, contextual branching and new-schema creation. Stability rules prevent isolated conflicts from prematurely displacing established states while version links preserve access to historical evidence. At inference time, retrieval follows these structures back to original dialogue. On a frozen 700-question LoCoMo subset, MINDSET achieves the highest observed answer F1 among six evaluated systems and significantly improves retrieval ranking. Without retuning, it achieves the highest judge accuracy on LongMemEval knowledge-update cases, compared with the other baselines evaluated. Additional experiments with GLM-4.7 and Gemma-4-31B demonstrate compatibility across answer backbones. These results support explicit, non-generative memory-state control as an effective approach to address evolving conversational knowledge.

cs.AI

新论文 · v1arXiv:2611.19841

Asking the right question is key to AI discovery in a chain of hypotheses

Hu Cang

原文摘要

Artificial intelligence (AI) scientists pursue discovery through a chain of hypotheses, fitting candidate models to existing observations and often choosing experiments where those models disagree most. Yet when early data miss a narrow feature, model selection favors simple curves that agree across the unmea- sured gap, so disagreement-guided search has no reason to look where the new phenomenon lies. When we tested two frontier AI scientists, Claude Sonnet 5.5 and GPT-6.1-Sol, across 32 closed-loop trials with a narrow peak omitted from the starting data, both sampled inside the peak region less often than random search. Surprisingly, the agent that followed model disagreement more closely missed the peak in all 32 trials, whereas discovery succeeded when experiments probed the unmeasured gap early and followed up across the anomaly. Asking the right question, rather than testing only where current models disagree, is therefore key to AI discovery in a chain of hypotheses.

cs.AI · physics.optics

版本更新 · v2arXiv:2605.15226

Is Agentic AI Ready for Real-World Hardware Engineering? A Deep Dive with Phoenix-bench

Qingyun Zou, Feng Yu, Hongshi Tan, Jiahao Cui, Bingsheng He, WengFai Wong

原文摘要

We ask whether agentic AI systems built for software engineering transfer to realistic hardware engineering. Existing hardware LLM benchmarks isolate sub-tasks but none jointly requires repository navigation, hierarchy-aware localization, Electronic Design Automation (EDA) executable verification, and maintenance-style patching. We introduce \textbf{Phoenix-bench}, a synchronized corpus of 511 verified Verilator instances from 114 GitHub repositories, each shipped with the developer patch, design-flow labels, fail-to-pass and pass-to-pass testbenches, and a Docker-pinned EDA environment so resolved-rate differences reflect agent behavior rather than toolchain availability. Using Phoenix-bench we run a uniform evaluation of four commercial agents and eight open-source agentic structures across four LLM backbones, plus two diagnostic interventions (file-level oracle localization and one round of testbench-log feedback). Three findings emerge. (i)~Software and hardware are fundamentally different engineering tasks: the same agent loses 37\% to 58\% from SWE-bench Verified to Phoenix-bench because hardware bugs propagate across parallel instantiated modules through signal flow rather than along a software-style call graph, and software-tuned agents stop at the symptom file instead of tracing back through the instantiation chain. (ii)~Failures concentrate on design control-flow / finite state machine (FSM) bugs, verification testbench bugs, and hard cases that demand cross-hierarchy signal-flow tracking and coordinated multi-file edits. (iii)~Localization granularity matters far more than localization itself: a perfect file-level oracle yields only $+1.4$\% because the agent then breaks files that did not need editing, while a single round of test case feedback lifts resolved rate by $42$\% to $45$\% because the test case tells \emph{where} the bug is and \emph{what} the fix has to look like.

cs.AR · cs.AI · cs.SE

版本更新 · v3arXiv:2605.12379

Discrete Flow Matching for Offline-to-Online Reinforcement Learning

Fairoz Nower Khan, Nabuat Zaman Nahim, Peizhong Ju

原文摘要

Many reinforcement learning (RL) tasks have discrete action spaces, but most generative policy methods based on diffusion and flow matching are designed for continuous control. Meanwhile, generative policies usually rely heavily on offline datasets and offline-to online RL is itself challenging, as the policy must improve from new interaction without losing useful behavior learned from static data. To address those challenges, we introduce DRIFT, an online fine-tuning method that updates an offline pretrained continuous time Markov chain (CTMC) policy with an advantage-weighted discrete flow matching loss. To preserve useful pretrained knowledge, we add a path-space penalty that regularizes the full CTMC trajectory distribution, rather than only the final action distribution. For large discrete action spaces, we introduce a candidate-set approximation that updates the actor over a small subset of actions sampled from reference-policy rollouts and uniform exploration. Our theoretical analysis shows that focusing on a small set of high probability actions can reduce computation while keeping the approximation error small, making our method practical for large action spaces. We evaluate on macro-action MinAtar games where offline-pretrained DRIFT outperforms its online-only counterpart across all five games and achieves the highest mean return on three and on oracle free Jericho, DRIFT achieves a mean normalized score of 7.3\% with a lightweight GRU encoder, outperforming the evaluated pretrained language-model baselines. Ablations demonstrate the importance of appropriately weighted path-space regularization and reference-guided candidate construction.

cs.LG · cs.AI

版本更新 · v2arXiv:2610.16275

A Survey of Flow Matching in Reinforcement Learning

Nabuat Zaman Nahim, Fairoz Nower Khan, Peizhong Ju

原文摘要

Flow Matching (FM) has recently emerged as a principled and efficient generative modeling framework for reinforcement learning (RL), enabling expressive, multimodal policy parameterizations via deterministic probability transport. Compared to diffusion-based policies that rely on stochastic denoising chains, FM uses sampling based on ordinary differential equations (ODEs), with learned velocity fields, which can substantially reduce inference latency and simplify the incorporation of RL objectives. As research in flow-based RL rapidly accelerates across offline continuous control, online fine-tuning, and foundation model alignment, the literature has become highly fragmented. In this survey, we provide a comprehensive taxonomy of flow-matching approaches in reinforcement learning. We organize the literature along two axes: the target distribution being modeled (e.g., action policies, value critics, transition dynamics) and the mechanism of RL signal integration (e.g., energy-weighted regression, flow-based policy gradients, and group relative policy optimization). Furthermore, we survey emerging frontiers such as discrete and non-Euclidean action spaces, provide a systematic comparative analysis against Gaussian and diffusion baselines, and outline critical open problems. Ultimately, this survey serves as a foundational roadmap for the next generation of generative reinforcement learning and alignment.

cs.LG · cs.AI

新论文 · v1arXiv:2611.19823

EfficientSkill: Learning Skills for Token-Efficient Multi-Turn Agents

Deep Shah, Sanket Badhe, Nehal Kathrotia, Priyanka Tiwari

原文摘要

Reflective skill optimizers improve multi-turn LLM agents by revising instructions from failed and successful rollouts, yet accumulating edge-case rules bloats prompts, inflates token use, and pushes agents into long, error-prone loops. We introduce EfficientSkill, a two-step framework that first discovers a reliable skill without token penalties, and then warm-starts accuracy-guarded pruning from that skill to compress instructions, bound agent turns, and offload reusable code into workspace utilities with zero implementation prompt cost. Across four frontier, mid-tier, and open-weight models on MATH-500, ALFWorld, and SpreadsheetBench, EfficientSkill cuts output tokens over unconstrained EvoSkill on 11 of 12 configurations (by up to 80%) while raising mean accuracy on every benchmark, including +14.3 and +17.5 percentage-point gains on SpreadsheetBench with Gemini 3.8 Flash and Grok 4.1 Fast by replacing stalled loops with compact workflows. By shrinking both generated tokens and replayed history, EfficientSkill lowers total token usage and serving cost across every evaluated configuration, with or without full-prefix KV caching.

cs.CL · cs.AI · cs.LG

新论文 · v1arXiv:2611.19822

The Flip Test: Video World Models Fail to Learn the World State

Mi Luo, Lijun Yu, Adams Wei Yu, Ang Cao, Baochen Sun, Ming-Hsuan Yang, Nan Hua, Yuan Hao 等 13 位作者

原文摘要

Video world models promise to predict the future of a scene from its present. Do current models keep this promise? We probe this question counterfactually. When a small change to the present flips the future, such as a box resting a few millimeters before or past a table edge, a model that tracks the state of the world must flip its prediction too. We build FlipTest, a three-level hierarchy of increasingly strict tests over 1,680 initial frames from 10 simulated and 15 real-world tasks. At the base, visual minimal pairs, two nearly identical frames whose true futures differ, test whether predictions change across the boundary; matched controls go further and test whether they change only where the future does; and dense sweeps, the strictest level, locate exactly where they change. All 11 video world models we evaluate fall short. The best averages 18.3% pair accuracy over real and simulated scenes, below the 25% of random guessing. Models flip as often on steps that leave the future unchanged as on steps that change it, and their response stays flat across the sweeps. On the same frames, a vision-language model is selective and traces a near S-shaped curve near the boundary, and a coding agent that measures the image comes close to the true boundary. This implies that the deciding information is in the frame, and video models just miss it. We therefore urge the community to rethink world modeling as pure video prediction. Next-frame prediction must do more than draw plausible pixels; it must also identify the causal states that control future dynamics.

cs.CV · cs.AI · cs.RO

新论文 · v1arXiv:2611.19821

GraphPass: A Study of Graph Neural Networks for Compiler Pass Selection

Mohammed Saqlain

原文摘要

Compiler pass selection is usually governed by fixed pipelines (-O2, -O3) or hand-written heuristics. We study whether a graph neural network (GNN) over a program's control-flow graph (CFG) can select among a fixed menu of five pass sequences in a single forward pass. GraphPass encodes each CFG with a graph attention (GAT) or graph convolutional (GCN) encoder and classifies it into one of the five sequences. On a synthetic benchmark of 5,000 generated CFGs in five program families, both encoders agree with the benchmark's reference label on about 92% of held-out programs, versus 68.4% for a hand-written rule-based heuristic and 85% for a flat MLP. An edge ablation indicates that this gain comes from per-block processing rather than from CFG edges: a GCN given the same basic-block features with all edges removed scores 95.0%, versus 92.2% with the true CFG. In this benchmark the label is determined by the program family, so the task amounts to family recognition. Real-compiler speedup is not measured. Inference costs O(|V| + |E|) per CFG, measured at 1.7 ms for a typical program and about 0.8 s for a 100,000-block graph on one CPU core.

cs.LG · cs.AI · cs.PL

新论文 · v1arXiv:2611.19812

Reasoning in the loop: antibody lead optimization by probing mechanism and measurement

Samuel Stanton, Aaron Solomon, Emily Delaney, Melanie Montgomery, Crystal Humphries, Timothy Acker, Ryan Henrici, Charbel Eid 等 10 位作者

原文摘要

Lab-in-the-loop systems, in which a model proposes molecules, automated assays measure their properties and the results retrain the model, are a popular approach to antibody lead optimization. These systems optimize for a specification defined with metrics computed from instrument signals, so their success depends on how well those metrics serve as proxies for the properties a drug candidate needs. When a metric is a poor proxy, neither a better model nor more data will fix it, and identifying the problem usually requires expert review. We report two campaigns run on BigHat Biosciences' Milliner platform, one on an anti-albumin nanobody (VHH) and one on an anti-mesothelin single-chain variable fragment (scFv), in which a research team and a general-purpose AI agent, Claude, brought scientific reasoning into the loop. Over seven rounds and about 1,010 design variants we built libraries from antibody lineage, structure and biophysics, and found from the raw curves that on each scaffold one thermal metric did not measure what its name implied. The libraries were built as controlled comparisons, so the results also show what each mutation contributes to each property. We fit a physical model of the curves, used it to choose metrics suited to each scaffold, and changed the experimental protocol, varying protein concentration and heating rate, to test the model against pre-registered predictions. Both leads improve on their parents in affinity, thermal stability and aggregation while improving or maintaining yield and purity.

cs.LG · cs.AI · q-bio.BM

版本更新 · v2arXiv:2611.02721

IssueLighter: An Issue Type-Aware LLM Agent for Issue Report Validity Assessment

Jiawei Ran, Yuxia Zhang, Ziyang Zhao, Hui Liu

原文摘要

Invalid issue reports are prevalent in open source software (OSS) projects, imposing substantial burdens on maintainers who already face a high risk of burnout. Prior research has primarily focused on identifying invalid reports for specific issue categories (e.g., bug and wontfix), leaving the broader landscape of OSS issue types largely unexplored and limiting the generalizability of existing approaches. Moreover, existing methods provide little support for explaining or validating their predictions, reducing their practical usefulness for maintainers. With the rise of LLM-based agents, it remains unclear whether a comprehensive understanding of the causes of invalidity across all issue types can improve validity assessment performance while producing well-grounded rationales. To this end, we first manually annotated the validity status of 520 issue reports from three OSS projects, i.e., pandas, scikit-learn, and mypy, resulting in 257 invalid reports and 263 valid ones. Then, we conducted a thematic analysis on these invalid reports and identified 13 causes of invalidity that are closely specific to issue types. Based on these insights, we propose IssueLighter, an agent-based framework with a type-aware validity assessment strategy that dynamically executes commands. Experimental results show that IssueLighter consistently and substantially outperforms all baseline methods, with an average F1 improvement of over 20 percentage points. Ablation studies confirm that both the type-aware strategy and the taxonomy of causes of invalidity contribute to the performance gains. Qualitative analysis further demonstrates that IssueLighter generates rationales grounded in accurately identified causes of invalidity and supported by evidence. Moreover, leveraging these rationales, we challenged maintainers' original decisions in four disputed cases, and all received approval.

cs.SE · cs.AI

版本更新 · v2arXiv:2606.05413

CausalPOI: Spatio-Temporal Graph-Based Causal Modeling for Cold-Start POI Check-in Forecasting

Zhaoqi Zhang, Miao Xie, Yi Li, Linyou Cai, Siqiang Luo, Gao Cong

原文摘要

As urban environments continue to evolve rapidly, accurately modeling the dynamic behaviour of Points of Interest is essential for supporting data-driven urban planning and commercial decision-making. While recent advancements in spatio-temporal graph learning have improved POI forecasting, most methods rely on proximity-based graphs and correlation-driven modeling, which overlook the functional dependencies between POIs and fail to capture the causal effects of urban interventions. In this paper, we introduce a novel research problem -- cold-start POI check-in forecasting, which aims to predict the future check-in pattern of a newly introduced POI, by modeling its temporal evolution and functional interactions with nearby POIs in a structured urban spatial context. To address these challenges, we propose CausalPOI, a spatio-temporal graph-based causal representation learning framework. CausalPOI leverages Spatio-Temporal Functional Interaction Graph to model semantic and spatial relationships between POIs, and constructs structurally aligned treatment and control graphs to simulate factual and counterfactual scenarios. Extensive experiments on real-world SafeGraph datasets demonstrate that CausalPOI significantly outperforms state-of-the-art baselines across the board, validating its effectiveness in spatio-temporal forecasting, semantic interaction modeling, and causal effect estimation, providing a more interpretable and actionable foundation for urban intervention analysis. Source code is available at Github.

cs.LG · cs.AI

版本更新 · v4arXiv:2603.24963

Template-Driven Machine Learning Standardization for Large Model Fleets

Jiang Liu, John Martabano Landy, Yao Xuan, Djordje Gligorijevic, Swamy Muddu, Nhat Le, Munaf Sahaf, Luc Kien Hang 等 23 位作者

原文摘要

Modern computational advertising platforms typically rely on recommendation systems to predict user responses, such as click-through rates, conversion rates, and other optimization events. To support a wide variety of product surfaces and advertiser goals, these platforms maintain an extensive ecosystem of machine learning (ML) models. Operating at this scale, however, creates significant development and efficiency challenges: substantial engineering effort is required to regularly refresh these models and propagate new techniques, resulting in long latencies when deploying innovations across the ecosystem. We present a large-scale empirical study comparing model performance, efficiency, and technique propagation between a standardized model-building approach and independent per-model optimization in recommendation systems. To facilitate standardization, we propose the Standard Model Template (SMT)---a framework that generates high-performance models adaptable to diverse data distributions and optimization events. By composing standardized ML model components, SMT reduces technique-propagation complexity from $O(n \cdot k)$ to $O(n+k)$, where $n$ is the number of models and $k$ the number of techniques. Evaluating an extensive suite of models over five global development cycles within Meta's production ads ranking ecosystem, our launched results demonstrate: (1) a 0.77\% average improvement in cross-entropy at neutral serving capacity; (2) a 1.52\% cumulative online metric lift; (3) a 92\% reduction in per-model iteration engineering time; and (4) a 6.3$\times$ increase in technique--model pair adoption throughput. These findings challenge the conventional wisdom that diverse optimization goals inherently require diversified ML model design.

cs.AI · cs.LG

新论文 · v1arXiv:2611.19807

MAPS: A Multimodal Agentic Framework for POI Data Production Based on Crowdsourced Street-view Imagery

Sijie Ruan, Tingyu Xie, Mingfei Liu, Yicheng Tang, Haoyu Han, Jingnan Qi, Jingshu Zeng, Dongqi Xu 等 10 位作者

原文摘要

Point-of-interest (POI) databases support location-based services, yet urban changes quickly make their records stale and manual maintenance is costly. Crowdsourced street-view images provide timely visual evidence, but converting them into database actions requires separating multiple POIs and irrelevant text in each image, then resolving incomplete observations against ambiguous records. We study Street-view Image-based POI Data Production and propose MAPS, the first framework to integrate structured POI extraction with database maintenance from such imagery. MAPS contains two cascaded modules. SPICE progressively adapts a lightweight vision-language model to reconstruct and filter text, group entries belonging to the same POI, and classify their semantics and completeness. RIMA retrieves candidate records, performs evidence-grounded POI matching and attribute-level reasoning, and selectively acquires complementary visual, geospatial, or external-source evidence when necessary. In human-reviewed full-task evaluation, MAPS outperforms the strongest baseline by 21.8\%. MAPS has also been deployed in Kuaishou's large-scale POI production pipeline, reducing average processing time by 76.9\% relative to manual annotation. These results demonstrate the effectiveness and practical value of MAPS for scalable POI maintenance.

cs.AI · cs.DB

版本更新 · v2arXiv:2607.29180

MoRAE: Flow-Friendly Self-Supervised Latents for Text-to-Motion Generation

Yifei Zhu, Dafei Qin, Zijiao Zeng, Mingyi Shi, Yangyang Cai, Taku Komura, Yoshifumi Kitamura

原文摘要

Which latent space should a text-to-motion generator model? We show that a self-supervised motion representation, once compressed, is a better generative space than latents learned by reconstruction. Our approach, MoRAE, is simple: we pre-train a Motion-JEPA encoder that predicts hidden joint trajectories from the rest of the body, and compress its frozen features into a compact latent for a standard flow transformer. Two observations drive this design. First, prediction yields features that capture action semantics and cross-joint structure better than reconstruction. Second, these features are ill-conditioned, and a standard flow transformer generates poorly in them directly. Compressing them into a compact latent makes generation reliable and places cross-joint structure on the directions the flow models most accurately. With a 58M DiT and the standard CLIP text condition, MoRAE achieves state-of-the-art FID on HumanML3D and KIT-ML while maintaining retrieval-based text consistency and diversity close to real motion.

cs.CV · cs.AI

版本更新 · v2arXiv:2609.19093

Reporting Practice Matters: The Impact of Reference Choice on Chest X-ray Report Evaluation

Daniel P. Jeong, Charles Q. Li, Hossein Hosseiny, Nitya M. Bhalla, Fatma Uyar Morency, Pradeep Ravikumar, Zachary C. Lipton, Michael Oberst

原文摘要

Radiologists follow heterogeneous reporting practices. Two radiologists examining the same image and identifying the same clinical findings might nevertheless compose superficially distinct reports, varying in terminology, shorthand, formatting, and level of detail. These variations in reporting norms represent an under-appreciated obstacle in efforts to evaluate AI-based radiology report generation (RRG) models, where machine-generated reports are typically assessed based on their concordance with human-generated references. In this paper, we quantify the sensitivity of established evaluation metrics to variations in reporting practices, revealing impacts large enough to alter the rankings of models. We introduce a radiologist-informed taxonomy of variations in radiology reporting practice and a method (ReRef) that rewrites reference reports along the axes of our taxonomy while preserving clinical interpretation. For instance, when comparing the performance of nine RRG models on MIMIC-CXR using RadCliQ-v1, condensing the discussion of normal findings in the reference reports causes Libra to drop from first to second place while CheXOne rises from third to first. Our results suggest that many current metrics fail to decouple clinical interpretation from conformity to reporting practices and that choosing the "right" references that accurately reflect the desired reporting practices can be important in practice. To support future research, we release MIMIC-CXR-Ext-ReRef, a radiologist-validated dataset of 120 (original, alternative) reference report pairs derived from MIMIC-CXR.

cs.CL · cs.AI

新论文 · v1arXiv:2611.19795

Lossless First-Action Menus for Linear-Reward Planning

Yukun Yang

原文摘要

We study sequential decision making where an agent reserves a minimum menu of initial actions before multi-objective trade-offs are revealed, followed by unrestricted continuation planning once linear reward weights arrive. We prove that a menu preserves the unrestricted optimal return across all linear scalarizations if and only if its reachable continuations cover all vertices of the full feature polytope. This geometric equivalence yields an output-sensitive algorithm that is fixed-parameter tractable in the vertex count, combining scalar backward induction with oracle-based facet identification. On explicitly layered graphs with unit feature increments, menu verification is coNP-complete and synthesis is complete for the second existential level of the polynomial hierarchy, even under additive loss strictly below one. In fixed dimensions two and higher, verification is polynomial whereas synthesis remains NP-complete, while one-dimensional instances admit polynomial-time solutions.

cs.AI

新论文 · v1arXiv:2611.19794

Eliciting Research Taste in LLMs through Future Research Direction Choice

Dhruv Trehan, Paras Chopra

原文摘要

Research taste is increasingly discussed in the context of AI scientists, yet remains little studied or measured in LLMs. This judgment matters especially in open-ended research, where a direction must be chosen before a verifiable objective is available. We operationalize research taste through one task of the scientific workflow: choosing which future research direction to pursue given a research goal and current results. Rather than predicting downstream scientometric metrics or treating author choices as objectively correct, we evaluate models against human-authored future-work directions and report author alignment, stable divergence, and unstable choice as behavioral measures of research-direction preference. We construct an order-controlled evaluation of 80 pairwise comparisons from 19 ML papers published between December 2025 and February 2026 across four subdomains. Each query presents a research goal, hypotheses tested with key results, and two candidate future directions: one author-recommended direction and one plausible distractor sampled from a similar paper. Across eight LLMs, author alignment ranges from 45.0% to 67.5%, while stable divergence ranges from 15.0% to 36.2%. These divergences largely persist under similarity controls, prompt ablations, and an author-preference framing check, and inter-model agreement suggests that some non-author preferences are shared across models. We characterize what models diverge toward using transition-category labels and selected reasoning examples, finding recurring movement toward scaling/generalization and trade-offs between methodological and infrastructure/data extensions. We view this as an early descriptive evaluation for eliciting research taste profiles in LLMs, and as a step toward understanding and steering their research preferences as they take on more autonomous roles in novel open-ended scientific discovery.

cs.LG · cs.AI · cs.CL

版本更新 · v2arXiv:2606.08969

CARE: A Conformal Safety Layer for Medical Summarization

Suhana Bedi, Bridget Lin, Anson Y. Zhou, Chloe O. Stanwyck, Jenelle A. Jindal, Sanmi Koyejo, David Stutz, Nigam H. Shah

原文摘要

Large language models (LLMs) are increasingly used for medical summarization but can omit medically important information and introduce unsupported claims. Existing error-detection methods produce heuristic or uncalibrated scores, providing no formal control over missed errors and no principled way to trade off safety against clinician review burden. We introduce Conformal Assessment for Risk Evaluation (CARE), a post-hoc risk-control layer that adds calibrated omission and hallucination annotations without retraining the summarizer. CARE uses scalar conformal risk control (CRC) to bound the probability that a document contains an unflagged hallucinated sentence, and Learn-Then-Test fixed-sequence testing (LTT-FST) to control the expected fraction of true omissions left unsurfaced for review. Omission surfacing depends jointly on sentence importance and summary coverage, a two-dimensional threshold-selection problem; we show that calibrating one gate and then imposing an uncalibrated second gate can invalidate risk control. Across five medical summarization tasks, mean held-out violation over 100 random calibration/test resplits remains at or below the target for both controllers. At $α=0.15$, LTT-FST achieves the lowest review workload on three of five datasets among omission baselines held to the same confidence guarantee. Review workload varies by task and is highest for long, high-compression documents. On 46 clinician-annotated documents, the calibrated hallucination controller flags 88% of clinician-identified hallucinated sentences. In a preliminary study of 75 clinician reviews, assisted review-time omission coverage increased from 50.4% to 79.0% without increasing review time. Overall, CARE provides a principled framework for targeted, risk-controlled review of LLM-generated medical summaries.

cs.CL · cs.AI

版本更新 · v2arXiv:2611.16748

PhysLoop: Harnessing LLMs' Physics Awareness for Self-Improving Video Generation Models

Yuxuan Chen, Xiaojie Li, Hao Chen, Juncheng Ma, Kewei Zhang, Zhongjing Du, Guoqiang Wei, Fangfang Wang 等 14 位作者

原文摘要

Video generation models produce visually realistic videos but still struggle with physical dynamics and causal consistency. Learning from generated videos requires fine-grained physical supervision, yet reliable quality judgments, correction pairs, and controlled variations are difficult to obtain. Large language models (LLMs) possess physical knowledge, but struggle to reliably guide video generation models in producing their training videos. We present PhysLoop, a unified harness designed to support physically grounded self-improvement in video generation models. The harness guides and constrains LLMs throughout video generation, selection, and correction, distilling their physical knowledge into physically consistent videos. The framework supports controllable training-data construction and inference-time video repair. By identifying and correcting failures, PhysLoop marks a key milestone toward recursive self-improvement (RSI) in video generation models with limited human intervention. We release 100K videos spanning diverse, realistic indoor and outdoor scenes, along with generation tools and two evaluation datasets: PhysLoop-Gen and PhysLoop-Repair. We evaluate controllable generation and localized repair on these datasets, respectively, and assess physical behavior on the external Physics-IQ Verified benchmark.

cs.CV · cs.AI

版本更新 · v2arXiv:2605.06219

Joint Consistency: A Unified Test-Time Aggregation Framework via Energy Minimization

Yunzhen Yao, Hongye Wang, Yahong Wang, Michael C. Gastpar, Bo Jiang, Lie He

原文摘要

This paper studies test-time aggregation, an approach that generates multiple reasoning traces and aggregates them into a final answer. Most existing methods rely on evaluation signals collected from candidate traces in isolation or answer frequencies, while ignoring comparative interactions among candidates. We propose Joint Consistency (JC), formulated as a constrained Ising-type energy minimization problem, which employs both independent evaluation signals and pairwise comparisons using LLM-as-a-judge. JC provides a unified framework for test-time aggregation that subsumes typical voting and weighted aggregation methods as special cases. Moreover, we develop an efficient approximation strategy that makes interaction modeling practical for large-scale test-time aggregation. Experiments on math and code reasoning benchmarks show that JC outperforms existing baselines across tasks, judge models, trace budgets, and trace-generation settings, with particularly pronounced gains on challenging problems.

cs.AI

新论文 · v1arXiv:2611.19779

Read Before You Fit: Text-Derived Constraints for Discovering Differential Equations from Noisy Biological Data

Pronoma Banerjee, Md. Masudur Rahman, Juan P. Wachs

原文摘要

Large language models are increasingly used for scientific equation discovery. However, most existing methods are strongly data-driven and assume sufficiently clean and representative observations. This assumption is problematic in biological systems, where individual-level measurements are often sparse, noisy, heterogeneous, and expensive to obtain. While the underlying mechanisms are extensively described in scientific text, they are not explicitly used for the LLM based modeling of such systems. We introduce P2C, an agentic framework that translates textual descriptions of biological systems into a typed set of formal constraints, and use these constraints to restrict candidate ordinary differential equation (ODE) structures before numerical fitting. P2C consists of five stages: Text2Constraint extraction, ODE discovery, symbolic verification, parameter fitting, and model scoring. We evaluate P2C on four published ODE systems of acute inflammation and immune regulation, under clean and noisy observations. By progressively degrading the numerical evidence, we isolate when textual domain knowledge contributes beyond regression alone. Across the four systems, adding Text2Constraint reduces the worst form error from 1.397 to 0.499 on clean data and from 13.096 to 1.095 for data perturbed with Gaussian noise of $σ=0.1$; P2C also achieves lower form error and trajectory error than several LLM-based baselines over the tested noise range.

cs.AI

版本更新 · v2arXiv:2608.09123

RISE-RL: Rubric-Informed Selective Exploration for Open-Ended Reinforcement Learning

Jinkun Hou, Zhuo Liu, Huimin Ren, Kuan Zhang, Zirong Chen, Hongsheng Xin, Pan Zhou, Kun Zhan

原文摘要

Aligning Large Language Models (LLMs) for open-ended tasks is challenging because responses must satisfy multidimensional criteria without following a single correct generation trajectory. Existing rubric-based reinforcement learning (RL) methods compress fine-grained criterion-level feedback into scalar rewards, making persistent capability gaps difficult to target under limited on-policy exploration. We propose $\textbf{RISE-RL}$ (Rubric-Informed Selective Exploration), which uses repeatedly missed rubric criteria to elicit privileged trajectories that are difficult to discover through unguided exploration alone. RISE-RL retains only trajectories whose complete-rubric reward exceeds the mean reward of natural rollouts, and then re-evaluates them under the original prompt to emphasize behaviors that remain weakly supported by the natural policy. The resulting guidance signal is optimized through a separate auxiliary objective and removed once its additional benefit diminishes. Experiments with 4B and 14B models across writing, chat, health, and science show that RISE-RL achieves the highest mean score on every evaluated benchmark under guidance-free evaluation. Compared with standard Rubric-RL, it improves the average score by 1.3 points at the 4B scale and $\textbf{3.3 points at the 14B scale}$, including a $\textbf{6.0-point}$ gain on CreativeWriting-V3. It also improves creative-writing diversity and yields gains on objectively scored medical and scientific benchmarks. These results indicate that selective internalization through reward filtering and policy support shaping is effective for open-ended reinforcement learning.

cs.AI

版本更新 · v6arXiv:2509.11080

Membership Inference Attacks on Recommender System: A Survey

Jiajie He, Xintong Chen, Wenjin Liu, Xinyang Fang, Min-Chun Chen, Yuechun Gu, Keke Chen

原文摘要

Recommender systems (RecSys) have been widely applied across e-commerce, finance, healthcare, and social media and have become increasingly influential in shaping user behavior and decision-making, underscoring their growing impact across domains. Since RecSys heavily relies on user data, its privacy concerns are significant and need to be addressed urgently. Recent studies on membership inference attacks (MIAs) in RecSys highlight this need. MIAs aim to infer whether a user or an interaction record was used to train a target RecSys model. Successful MIAs can lead to severe privacy breaches, such as revealing sensitive aspects of users' lifestyles. MIAs in RecSys have features distinct from those targeting classification models or large language models. However, no systematic survey on this topic has yet been conducted. We present the first comprehensive survey on RecSys MIAs, exploring their taxonomy, design principles, evaluation methods, and defense mechanisms. Based on the summary of existing studies in this area, we also outline several promising future research directions. This survey aims to raise awareness of privacy risks among RecSys researchers, practitioners, and users, and promote privacy protection practices in RecSys design.

cs.IR · cs.AI · cs.CR

版本更新 · v2arXiv:2611.07604

FisherDistill: Diffusion Distribution Distillation via Closed-Form Fisher Critics

Mahesh Bhosale, Yuanhao Zhai, Siwei Lyu, Junsong Yuan, David Doermann, Defang Chen

原文摘要

Recent advances in diffusion distillation have enabled high-quality image generation in only one or a few network function evaluations (NFEs). Despite their efficiency, aggressively distilled generators often retain a residual distribution gap from their multi-step teachers, motivating recent methods to introduce adversarial refinement with a learned discriminator. In this work, we ask whether such refinement necessarily requires a learned adversarial discriminator. We introduce \textbf{FisherDistill}, a simple alternative based on Fisher linear discriminant analysis. Using intermediate representations from the frozen teacher, we estimate the first- and second-order statistics of real and generated samples and derive a closed-form Fisher direction that maximally separates their feature distributions relative to within-class variation. This direction provides a lightweight refinement signal that complements the original distillation objective without requiring an additional adversarial discriminator or its optimization loop. Across multiple distillation frameworks, our approach consistently improves generation quality over their respective distilled baselines and substantially closes the gap to adversarially refined counterparts. Code: https://bhosalems.github.io/FisherDistill/

cs.CV · cs.AI

新论文 · v1arXiv:2611.19753

Learning Autoregressive Neural Operators on Solver-Defined Physical Constraint Manifolds

Nannan Zhou, Guanmeng Xian, Ning Yang

原文摘要

Autoregressive neural operators approximate time-dependent PDE solvers by repeatedly composing a learned transition, although the resulting states need not remain feasible under the discrete constraints of the numerical problem. We study this issue for incompressible flow, where each velocity state must satisfy solver-level incompressibility together with prescribed boundary values. We formulate autoregressive operator learning as a map between solver-defined affine feasible sets and introduce PCM-NO, a projected-residual realization of this feasible-transition principle. The neural backbone predicts a candidate state increment, a fixed tangent projection extracts its admissible component, and a metric retraction maps the advanced state to the prescribed feasible set. With exact affine constraint solves, the resulting transition is closed on the feasible set, while an affine identity separates admissible-update parameterization from state-level closure. With approximate solves, tangent-only residuals admit a cumulative upper bound, whereas per-step retraction resets feasibility to the residual achieved by the current solve. Experiments on periodic Navier--Stokes flow and lid-driven cavity flow demonstrate the superiority of PCM-NO in preserving solver-level constraints over long autoregressive rollouts and under parameter, boundary-value, and resolution shifts, while achieving lower rollout errors and better flow-specific diagnostics across both in-distribution and stress settings.

stat.ML · cs.AI · cs.LG · physics.flu-dyn

新论文 · v1arXiv:2611.19752

StaleBench: Isolating Stale Retrieval in Agent Memory, and Pricing the Fix

S M Mahbub Murshed

原文摘要

An agent that stores facts over time eventually stores several versions of the same fact. Asked a value-agnostic question - "which team does X play for now?" - a similarity retriever cannot prefer the current version, because every version matches the entity the query names. StaleBench isolates this stale-retrieval failure on 892 real updating facts, with currency labels from the source (Wikidata validity years via TempLAMA), never from a retrieval score. Our central result is that retrieval metrics can rank memory policies backwards: a retention policy and its retain-everything control return the identical statement at rank 1 for every test query, yet differ by 29.5 and 65.6 points in reader accuracy on two readers when no date is shown. Contamination@5, the metric we add, ranks them correctly where current@1 and current@5 do not. Withdrawing the lexical anchor that statements carry costs deployed keying mechanisms tens of points of freshness; entity resolution on both the write path and the query recovers most of the loss. Three deployed systems (Mem0, Graphiti, LangMem) and a reimplemented ERASE each handle supersession differently, and no single scoring convention describes all four. We release the data, a frozen evaluator, and reference baselines.

cs.CL · cs.AI · cs.IR

新论文 · v1arXiv:2611.19747

How Far Is Too Far? Object Recognition Declines Monotonically with Semantic Distance

Domenic Bersch, Hai Van Tran, Gemma Roig

原文摘要

Vision models are sensitive to object-background context, yet it remains unclear how recognition changes as the semantic distance between objects and backgrounds increases. We introduce ImageNet-OOC1k, a benchmark enabling graded control of object-background relationships along a continuous consensus visual-semantic dissimilarity axis. This formulation reveals a highly consistent monotonic decline in recognition accuracy across 35 models spanning convolutional neural networks, Vision Transformers, and vision-language models. Internal analyses of Vision Transformers reveal patterns consistent with aggregation failure, where object evidence is present at the patch level but not consolidated into the final prediction. Using this characterization as a diagnostic probe, a simple spatial re-ranking strategy recovers a subset of these errors and improves in-sample accuracy by up to 5.7 percentage points. These results establish context sensitivity as a continuous and predictable property of modern vision systems, with Vision Transformer analyses suggesting that aggregation limitations may contribute to some failure types even when object evidence remains available.

cs.CV · cs.AI

新论文 · v1arXiv:2611.19742

Exploration of Reinforcement Learning in Generative Recommendation Models

Pengfei Zheng, Wuchao Li, Shiyao Wang, Honghui Bao, Hongyi Xie, Jiaxin Deng, Xingmei Wang, Jinghao Zhang 等 14 位作者

原文摘要

Traditional cascaded recommender systems balance accuracy and efficiency over large item spaces, but misaligned stage-wise objectives hinder end-to-end optimization. Generative recommendation unifies these stages in an autoregressive framework, providing a natural interface for reinforcement learning (RL) to optimize the entire generation process. However, directly applying RL often leads to unstable optimization and degraded recommendation quality. We revisit Group Relative Policy Optimization (GRPO) and other ratio-based objectives through a unified logit-gradient view. We identify three issues: (i) clipped surrogates create dead-gradient regions, under-utilizing samples; (ii) negative-advantage updates can become effectively unbounded as importance ratios grow, allowing a small subset of negatives to dominate optimization; and (iii) ratio-based control lacks sensitivity to absolute policy confidence, so updates on high-confidence cases may fail to vanish. Motivated by this diagnosis, we propose \emph{Gradient-Bounded Policy Optimization} (GBPO), a simple objective with stop-gradient normalization that is simultaneously (a) fully gradient-active, (b) sign-symmetrically bounded for both positive and negative updates, and (c) explicitly confidence-weighted under current policy probabilities. This ensures that updates gracefully vanish for $π\!\to\!1$ positives and $π\!\to\!0$ negatives. Across the evaluated settings, GBPO achieves competitive recommendation performance with more stable training than competing RL methods.

cs.LG · cs.AI