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

+197新增论文
+151其中 AI 相关
+16已有论文修订
+91新增 AI 导读

arXiv 最近完成扫描:2026-11-21 12:19 · 全站最近入库:2026-11-21 12:19各信源进度 →

统计口径与数据新鲜度

统计 2026-11-21 05:44 — 2026-11-22 05:44(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:2026-11-21 12:19最近发现新增版本:2026-11-21 12:19库内论文最新更新时间:2026-11-20 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。

29,158 已收录论文24,181 AI 相关8,992 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 19 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-21 12:17(北京时间)
新论文 · v1arXiv:2611.26120

Better Retrieval, Not Better Binding: Pretraining Exposure in Egocentric Video Retrieval

Yuqi Guo, Michael Wray

原文摘要

Video-language models (VLMs) achieve strong performance on video-text retrieval and are increasingly used for egocentric video understanding. Yet, despite gains in zero-shot retrieval, they have shown limitations in compositional understanding, and it remains unclear whether they understand compositional relationships between verbs and nouns. In this work, we introduce Ego-ExBind, a compositional diagnostic protocol for analyzing pretraining exposure and verb-noun binding in egocentric video-text retrieval. Ego-ExBind consists of an Exposure Ledger built from the EgoClip pretraining corpus used by EgoVLPv2, which defines Seen Composition, Unseen Composition, and Unseen Atom splits for EPIC-KITCHENS-100 multi-instance retrieval (MIR). We further probe noun and verb binding channels through controlled noun-swap and verb-swap negatives. We find that zero-shot retrieval is positively associated with pair exposure within seen compositions, even after controlling for the marginal frequencies of verbs and nouns. However, higher pair exposure is associated with larger noun-swap margins and slightly smaller verb-swap margins. This asymmetry motivates us to test whether exposure reweighting can improve retrieval with little change in the measured swap margins. To isolate the effect of exposure input, we keep the trained checkpoint fixed and replace only the exposure input at inference. We demonstrate that this exposure reweighting can improve video-text retrieval without stronger verb-noun compositional understanding, and that the gain depends on the correct sample-level exposure correspondence. By separating exposure-driven retrieval gains from verb-noun binding, Ego-ExBind provides a diagnostic framework for evaluating whether future improvements in egocentric MIR reflect compositional understanding rather than pretraining familiarity.

cs.CV

版本更新 · v2arXiv:2609.12157

Direct Topology Tracking in Continuous Implicit Models

Guanqun Ma, David Lenz, Kaiyuan Tang, Hanqi Guo, Chaoli Wang, Tom Peterka, Bei Wang

原文摘要

We present a framework for tracking topological features directly within continuous implicit models. Such models, including implicit neural representations (INRs) and multivariate functional approximations (MFAs), are increasingly adopted to represent scientific data without the resolution constraints of discrete grids. They offer compact, smooth, and differentiable representations of complex fields, enabling new opportunities for high-performance data storage, reconstruction, and analysis. Given a continuous implicit model, our method tracks the evolution of critical points by querying the model and its derivatives, thereby eliminating the need to resample onto a grid. This approach enables faithful feature tracking while avoiding discretization-induced artifacts such as aliasing. We demonstrate the generality of our framework across a range of implicit representations, including analytic functions, MFAs, and INRs, and show that it produces smooth, coherent critical point trajectories. By enabling feature tracking directly on continuous representations, our method supports a new class of feature-driven visualization workflows centered on implicit models.

cs.CG · cs.LG

版本更新 · v2arXiv:2604.05375

DAT: Dual-Aware Adaptive Transmission with Small-Large Model Collaboration for Edge-Cloud Anomaly Detection

Qi Guo, Zheming Yang, Yunqing Hu, Chang Zhao, Wen Ji

原文摘要

Continuous video anomaly detection is critical for timely safety monitoring in edge--cloud visual systems, but it remains challenging because abnormal events are sparse, transient, and require both low-latency anomaly detection and verifiable visual evidence. Multimodal large language models (MLLMs) offer strong semantic understanding and visual reasoning capabilities for such event-centric analysis, yet applying them to continuous video streams incurs prohibitive computation and communication overhead under bandwidth-constrained edge--cloud deployments. To address this challenge, we propose DAT, a dual-aware adaptive transmission framework that supports high-quality semantic generation, low-latency anomaly detection, and progressive visual evidence supplementation. DAT first uses a collaborative small--large model cascade, where a lightweight edge-side model filters non-target-event frames, performs object detection, and triggers MLLM inference only for suspicious frames. Building on this cascade, DAT introduces visual guidance and semantic prompting to improve structured event understanding, localization, and output consistency. It further schedules semantic anomaly alarms and visual evidence through a semantic- and bandwidth-aware multi-stream transmission optimization method. Experimental results show that DAT achieves 97.77\% F1 in traffic-accident detection while substantially reducing inference and communication overhead. On continuous video workloads, DAT lowers the MLLM call rate to 13.71\% and reduces upload volume by at least 95.72\% compared with representative edge--cloud baselines. Under severe congestion, DAT reduces weighted alarm delay by up to 87.9\%, demonstrating efficient MLLM-based anomaly detection under constrained edge--cloud networks.

cs.MM

版本更新 · v2arXiv:2610.29691

Efficient Generative Image Compression with Quantized Latents Refinement

Hao Cao, Yinhuan Huang, Wenqi Guo, Yuan Shen, Jungong Han, Zhijin Qin

原文摘要

Generative image compression (GIC) delivers strong perceptual quality. However, GIC methods with strong perceptual quality often rely on pretrained diffusion models with high latency, while recent lightweight GIC methods reduce latency but compromise perceptual quality. At low bitrates, coarse deterministic quantization can map nearby encoder latents representing different fine details to the same quantized point, making those details difficult to recover. Motivated by this, we propose GR-GIC which delivers competitive compression performance with fast inference. For better performance, we introduce Guided Refinement (GR), which combines latent refinement guided by entropy context with adaptive perturbations, jointly improving reconstruction at low bitrates. For fast inference, we introduce reorganized projected residual vector quantization, which equivalently reorganizes residual updates into low-dimensional lookups to expose greater parallelism and reduce latency. Experiments on Kodak show that GR-GIC achieves BD-rate savings of 22.09%, 42.14%, 19.35%, and 30.87% over OneDC for LPIPS, DISTS, PSNR, and FID, respectively. Moreover, our CUDA implementation encodes and decodes in 2.81 and 3.48 ms on an NVIDIA A100, achieving $27.36\times$ and $58.77\times$ speedups over the official OneDC implementation, respectively. We further evaluate GR-GIC on an NVIDIA Jetson Orin NX, demonstrating competitive coding speed on edge hardware.

cs.CV

版本更新 · v3arXiv:2603.21465

DRTriton: Large-Scale Synthetic Data Driven Reinforcement Learning for Triton Kernel Generation

Siqi Guo, Ming Lin, Tianbao Yang

原文摘要

Developing efficient CUDA kernels is a fundamental yet challenging task in the generative AI industry. Recent research leverages Large Language Models (LLMs) to automatically convert PyTorch reference implementations to CUDA kernels, significantly reducing engineering effort. State-of-the-art LLMs, such as GPT-5.2 and Claude-Sonnet-4.5, still struggle with this task. To address this challenge, we propose DRTriton, a scalable learning framework for training LLMs to convert PyTorch programs into highly optimized Triton kernels, which are then compiled to CUDA kernels at runtime. DRTriton consists of three key components: (i) a data synthetic algorithm CSP-DAG that guarantees full coverage and unbiased uniform sampling over the operator space with controlled difficulty; (ii) a curriculum RL framework with decoupled rewards that jointly optimizes conversion success rate and execution speed; and (iii) a test-time search algorithm that further improves the execution speed of the generated Triton kernels. With a warmup stage of SFT on limited PyTorch-Triton pairs curated using existing LLMs, DRTriton trained by RL on synthesized PyTorch programs generalizes effectively to real-world CUDA kernels that are challenging even for human experts. Experimental results show that DRTriton-7B achieves speedup over PyTorch on 92% of KernelBench Level 2 tasks, compared to 23% for GPT-5.2 and 19% for Claude-Sonnet-4.5.

cs.CL · cs.LG

版本更新 · v2arXiv:2610.41216

TinyAstral:将自进化执行框架迁移至小型通用操作智能体

TinyAstral: Transferring Self-Evolving Harnesses to Tiny Generalist Manipulation Agents

Kehong Gong, Xiaoxu Zheng, Guanli Hou, Seunghun Lee, Mengqi Guo, Runjia Tan, Zhengyu Wen, Qi Wang 等 11 位作者

AI 速读 基于摘要

先通过递归自我改进进化大型教师模型的执行框架,再将其蒸馏到4B小型学生模型,使紧凑操作智能体在RoboTwin等任务上接近教师表现并降低推理成本。

cs.RO · cs.CV

查看原文摘要

Robot agents built on vision-language models depend on both the planner and its harness: the tools, feedback, memory, and recovery logic around it. Can improvements to a large teacher's harness be carried into a compact agent with lower inference compute cost? We present TinyAstral, a two-stage approach. We first freeze an open-weight teacher and improve its harness through recursive self-improvement (RSI): frontier LLM agents read the teacher's trajectories and propose harness edits, and each edit is accepted only if it passes a paired statistical test. We then freeze the final harness, collect teacher trajectories with reasoning and failure recovery, and distill a small student that replaces the teacher in the same harness. On RoboTwin, harness evolution raises the teacher's success from 33.4% to 63.6%; the final 4B student reaches 62.8%. Across four harness versions, distilled 4B students score close to their corresponding teachers, and later versions yield stronger students, showing that harness evolution gains carry into compact models. The harness gains also transfer to other frozen planners without retraining. On identical episodes, the final 4B student outperforms open 4B agents, the open 27B harness-RSI model SenseNova-RoboRSI, and our own teacher model placed in three competitors' harnesses. It stays close to its teacher on all three training benchmarks, and on three benchmarks never used for RSI or training it scores close to a 552B planner running the same harness. On the measured deployment episodes, it uses about one fifth of its teacher's accelerator time. Project page: https://garfield-kh.github.io/TinyAstral/

版本更新 · v2arXiv:2508.05167

PhysPatch: A Physically Realizable and Transferable Adversarial Patch Attack for Multimodal Large Language Models-based Autonomous Driving Systems

Qi Guo, Xiaojun Jia, Shanmin Pang, Simeng Qin, Lin Wang, Ju Jia, Yang Liu, Qing Guo

原文摘要

Multimodal Large Language Models (MLLMs) are becoming integral to autonomous driving (AD) systems due to their strong vision-language reasoning capabilities. However, MLLMs are vulnerable to adversarial attacks, particularly adversarial patch attacks, which can pose serious threats in real-world scenarios. Existing patch-based attack methods are primarily designed for object detection models and perform poorly when transferred to MLLM-based systems due to the latter's complex architectures and reasoning abilities. To address these limitations, we propose PhysPatch, a physically realizable and transferable adversarial patch framework tailored for MLLM-based AD systems. PhysPatch jointly optimizes patch location, shape, and content to enhance attack effectiveness and real-world applicability. It introduces a semantic-based mask initialization strategy for realistic placement, an SVD-based local alignment loss with patch-guided crop-resize to improve transferability, and a potential field-based mask refinement method. Extensive experiments across open-source, commercial, and reasoning-capable MLLMs demonstrate that PhysPatch significantly outperforms prior methods in steering MLLM-based AD systems toward target-aligned perception and planning outputs. Moreover, PhysPatch consistently places adversarial patches in physically feasible regions of AD scenes, ensuring strong real-world applicability and deployability.

cs.CV

版本更新 · v3arXiv:2608.23018

SplitLite: Low-Rank Residual Compression for Split Learning

Tao Li, Yulin Tang, Qi Guo, Xianhao Chen

原文摘要

Federated fine-tuning of on-device large language models (LLMs) faces a significant computing burden. To overcome this limitation, split learning (SL) has emerged as a promising solution, which offloads the primary training workload to a powerful server. However, SL requires exchanging high-dimensional activations and gradients between clients and the server, resulting in prohibitive communication costs. To overcome this challenge, we propose SplitLite, a communication-efficient split federated LoRA fine-tuning method that exploits the low effective rank structure of consecutive-epoch activation and gradient residuals. Our key finding is that, under low-rank LoRA updates with the pretrained backbone frozen, the activation and gradient residuals of the same data sample between adjacent epochs exhibit low effective ranks, with most of their energy concentrated in a small number of singular directions. By exploiting this property, SplitLite transmits only quantized truncated singular value decomposition (SVD) residual factors, thereby significantly reducing both activation uplink and gradient downlink traffic. Extensive experiments on the GLUE benchmark across a series of advanced on-device LLMs demonstrate that our method reduces activation uplink communication costs by up to 93.5\% and total communication costs by up to 88.6\%, without sacrificing model performance.

cs.LG · cs.AI

版本更新 · v2arXiv:2610.11316

MetaEncoder: Exploring the Limit of Bi-Encoders for Multimodal System One Decision Making with Natural Language Interface

Jianpeng Cheng, Guangyu Sun, Aashu Singh, Benyu Zhang, Haixing Dai, Hossein Mansour, Jiangfan Zhang, Shlok Kumar Mishra 等 14 位作者

原文摘要

System One models prioritize fast, low-latency decision-making and calibrated probability distributions over token generation. While prevailing systems rely on structured schemas and typed outputs, candidate-scoring trade-offs remain poorly understood. We systematically investigate the scaling dynamics, accuracy, and inference efficiency of three scoring architectures adapted from a pre-trained decoder: bi-encoders, cross-encoders, and constrained decoders. Experiments show clear performance scaling with model size across all three architectures. Cross-encoders consistently achieve the highest accuracy across model sizes, while bi-encoders and constrained decoders perform similarly as training tokens scale. Efficiency-wise, bi-encoders maintain near-constant inference latency as candidate pools grow because candidate embeddings can be cached offline, whereas cross-encoder and constrained-decoder costs scale quadratically with candidate size. This efficiency advantage makes bi-encoders uniquely suited to bridge small closed-set decisions and massive open-set retrieval. For closed-set tasks, including all candidate options in the request prompt is critical for bi-encoder competitiveness. Guided by these insights, we introduce MetaEncoder, a multimodal encoder fine-tuned from a Muse-Glimmer decoder via unidirectional contrastive learning to match requests and candidates. We present packed contrastive training to maximize efficiency for dynamically allocated negative candidates, enabling memory-efficient full fine-tuning of a 30B decoder into an encoder. Evaluated across 12 benchmark suites spanning closed-set decision-making, multimodal understanding, and open-set retrieval, MetaEncoder outperforms state-of-the-art multimodal embedding baselines on decision tasks and MMEB-v3 image/video retrieval, while scaling to open candidate spaces far too large for current decision models.

cs.CL · cs.AI

新论文 · v1arXiv:2611.05994

FedIA: Instance-Aware Cross-Modal Representations for Federated Prompt Learning

Xuanjing Zhang, Qi Guo, Yanan Qiao, Yong Han, Junge Bo

原文摘要

Federated Prompt Learning (FPL) enables parameter-efficient adaptation of pre-trained Vision-Language Models across decentralized clients without exchanging raw data. To cope with cross-client heterogeneity, existing FPL methods typically pair a globally shared prompt with a client-specific local prompt. However, this design personalizes only at the client level: the same local prompt is applied to all samples of a client, and the global and local prompts are combined in an input-agnostic, fixed manner. Consequently, it overlooks the substantial visual variations among samples within the same client and cannot adjust the balance between shared and personalized knowledge to each input. To address these limitations, we propose Federated Prompt Learning with Instance-Aware Cross-Modal Representations (FedIA), an FPL framework that learns instance-aware cross-modal representations. Specifically, FedIA first uses global text prompts as client-invariant semantic anchors to select image patches most relevant to each candidate class, suppressing distracting regions. The selected visual evidence then conditions the generation of instance-aware local text representations. Finally, an adaptive global-local fusion module calibrates the contributions of local and global text representations for each image-class pair. In this way, FedIA shifts personalization in FPL from the client level to the instance level while preserving the semantic knowledge shared across clients. Extensive experiments under domain shift, pathological label skew, and Dirichlet-based label skew show that FedIA outperforms FPL baselines, with maximum accuracy gains of 1.44\%-4.37\% over the latest baseline across the three settings. Ablation studies and analyses further verify the complementary contributions of instance-aware representation generation and adaptive global-local fusion.

cs.CV · cs.LG

新论文 · v1arXiv:2611.04556

Inter-Agent Influence: Evaluating Persuasion, Deception, and Coercion in Multi-Agent Systems

Chandler Smith, Qi Guo, Cecilia Elena Tilli, Sophia Hatz, David Demitri Africa, Patricia Paskov, Christian Schroeder de Witt, Philip Torr 等 9 位作者

原文摘要

The deployment of AI agents in multi-agent settings enables inter-agent influence, whereby agents can strategically steer each other's behaviour towards an intended outcome. Such capabilities pose novel risks, as malicious agents could direct others towards harmful actions. In this work we investigate three such capabilities - persuasion, deception, and coercion - across five realistic evaluation environments and a variety of agents instantiated with frontier language models. We observed significant inter-agent influence capabilities; in oversight environments, all tested models shifted overseer decisions from rejection to approval through persuasion, deception, or coercion. We also observed successful influence in non-hierarchical 'peer-to-peer' settings through redirection of a peer's safety research trajectory. As multi-agent deployments become increasingly commonplace, these results point to a pressing need for work on mitigating the risks from inter-agent influence. Code and data are available at https://github.com/Cooperative-AI-Foundation/inter-agent-influence.

cs.AI · cs.MA

新论文 · v1arXiv:2610.40207

RoboAtom: Large-scale Atomic-Action Supervision for Robot Learning

Xueyuan Wei, Xuanxu Wang, Xudong Li, Yuping He, Wanshui Gan, Jiaqi Peng, Wenqi Guo, Jianjun Zhou 等 13 位作者

原文摘要

Fine-grained action annotations can support skill reuse and compositional generalization in robot policies. However, robot datasets typically provide task-level instructions without subtask annotations. Although some include subtask labels, inconsistent formats and conventions hinder cross-source joint training. We propose RoboAtom-Data, which adds unified, time-aligned atomic-action annotations to existing data from multiple sources, covering 1.45 million trajectories, 6.88 million skill-labeled segments, and 16,882.1 hours. Our pipeline combines a unified protocol, large-scale manual annotation, propagation across repeated executions with the same ordered skill sequence, and multi-stage quality control. We evaluate the dataset through automatic atomic-action annotation and robot control. We construct RoboAtom-Bench to standardize evaluation of skill recognition and temporal boundary localization. Trained on the annotated data, RoboAtom-Anno achieves better automatic annotation performance than all evaluated open- and closed-source models on both RoboAtom-Bench and the external WGO-Bench. For robot control, atomic-action supervision paired with hierarchical execution improves task success over task-level baselines on LIBERO-Long, LIBERO-Plus, and RoboTwin. We will release the data, benchmark, model, and construction pipeline to support further research on skill-level robot learning.

cs.RO

版本更新 · v2arXiv:2610.34239

Honesty Testing of LLMs and AI Agents: A Vision for Software Engineering

Ying Xiao, Xinjian Wu, Lukas Twist, Yaoqi Guo, Yepang Liu, Zhenpeng Chen, Jie M. Zhang

原文摘要

Human oversight of large language models (LLMs) and artificial intelligence (AI) agents increasingly depends on their reports. Claims that conflict with a system's own judgments can conceal failures, prevent timely intervention, and contribute to severe harm. Greater autonomy makes faithful reporting increasingly important for maintaining human control. In this vision paper, we propose honesty testing of LLMs and AI agents as a new research area in software engineering. Building on existing honesty evaluations, our goal is to make faithful reporting a software quality requirement that developers can specify, test, and maintain. We define honesty bugs as behavioural defects in which an LLM or AI agent, when required to communicate faithfully, makes an explicit claim that conflicts with its own judgment at the time of reporting. We present an initial framework covering LLM-level testing under controlled inputs and agent-level testing of complete systems. The framework connects these levels to testing workflows and oracle evidence. We identify research directions in benchmarks, model and harness analysis, oracle construction, and lifecycle testing to support effective oversight alongside task correctness.

cs.SE

版本更新 · v2arXiv:2610.34817

Silent and Costly: Characterizing and Detecting Prompt-Caching Mis-Integration in LLM Agents

Ying Zhang, Ruoqi Guo, Yi Liu, Gelei Deng, Yuekang Li, Yepang Liu, Leo Zhang

原文摘要

Large language model (LLM) services process massive volumes of user requests, resulting in substantial inference costs. To reduce both cost and latency, providers have introduced prompt caching. However, correctly integrating prompt caching is non-trivial due to heterogeneous provider-specific APIs and continuously evolving caching semantics. To understand prompt-caching mis-integration in downstream applications, we conduct the first empirical study of 120 prompt-caching-related pull requests from 10 popular open-source agent systems. Our findings show that prompt-caching mis-integrations arise in both request-side misconfiguration and read-side usage miscount. More importantly, most mis-integrations fail silently without causing observable runtime failures. Motivated by these findings, we design CACHELINT, which models a prompt-caching mis-integration as an invocation site together with the preconditions required for correct cache integration. CACHELINT locates caching invocation sites with CodeQL and discharges each precondition either against a dated provider capability oracle or with an LLM guard check. CACHELINT achieves 0.71 precision and 0.78 recall in identifying the 120 labeled mis-integrations across the 10 studied agent systems. Our evaluation further demonstrates the effectiveness of CACHELINT in real-world settings, identifying 32 mis-integrations across 571 agent projects with 0.86 precision. These findings have received 21 developer confirmations and led to 18 upstream fixes, including one in Anthropic's official SDK.

cs.SE · cs.AI

版本更新 · v2arXiv:2608.00643

Domain Decoupling Attack: Exploiting the Validation Gap Between Protective DNS and Shared Edge Routing

Weizhe Wang, Minhong Dong, Jinhao Li, Lyuchuan Wei, Yao Zhang, Qiang Hu, Qi Guo, Guangquan Xu 等 9 位作者

原文摘要

Command-and-control, data transfer, and censorship circumvention can hide in traffic to widely used services. DNS-bound access control grants permits based on resolved IP addresses, while shared edges route requests by tenant identity, creating a validation gap. An attacker can resolve an allowed anchor, obtain a permit for its shared edge address, and access a hidden tenant by presenting the tenant's name as both SNI and HTTP Host. To investigate this gap, we present the Domain Decoupling Attack (DDA), which exploits this gap in CDN and non-CDN hosting while preserving SNI--Host consistency. Across 1,069,048 domains, we conduct 18,025,068 successful probes. Overall, 95.8% of domains accept an external SNI, and 84.6% return content matching the named tenant. Exposure reaches 99.26% across 506,763 CDN domains and 92.75% across 562,285 non-CDN domains. The co-location graph places 666,413 exposed domains in cross-tenant components; 97.7% of 54,069 cross-tenant non-CDN addresses have a confirmed path to another tenant. None of the tested defenses bind tenant identity to the DNS-approved name. Cleartext inline IPS can block access based on SNI, but Encrypted Client Hello (ECH) hides it. The Windows Zero Trust DNS (ZTDNS) permit remains reusable in both cases, enabling the concealment of command-and-control and data-exfiltration traffic and facilitating censorship circumvention. The risk conceals command-and-control and data-exfiltration traffic and supports censorship.

cs.CR · cs.NI

新论文 · v1arXiv:2610.34573

Software Engineering for AI-Driven Science

Lukas Twist, Xinjian Wu, Jingzhi Gong, Shuyin Ouyang, Ying Xiao, Benedetta Donato, Yaoqi Guo, Zhenpeng Chen 等 11 位作者

原文摘要

AI has made significant contributions to scientific discovery. Autonomous agents can now independently analyse complex problems, design experiments, write and execute code, analyse findings, and derive scientific conclusions. Nevertheless, errors in AI-generated code, experimental procedures, or evidence interpretation can lead to invalid research conclusions, undermining the reliability and reproducibility of automated scientific discovery. In this vision paper, we propose Software Engineering for AI-Driven Science (SE4AI4Sci) as an emerging research area for addressing this challenge. As an organising framework, we introduce the Scientific Assurance Chain, connecting intent, experimental design, implementation, execution, evidence, and claim. Through six documented failures in recent AI-for-science systems, we show where this chain can break, where existing assurance mechanisms fall short, and what new research challenges follow. We also outline a research agenda for adapting and extending software engineering techniques across the chain.

cs.SE

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1334ACL Anthology · ACL 2026

N-GLARE: An Non-Generative Latent Representation-Efficient LLM Safety Evaluator

Zheyu Lin, Jirui Yang, Yukui Qiu, Yubing Bao, Hengqi Guo, Yao Guan

原文摘要

Evaluating the safety robustness of LLMs is critical for their deployment. However, mainstream Red Teaming methods rely on online generation and black-box output analysis. These approaches are not only costly but also suffer from feedback latency, making them unsuitable for agile diagnostics after training a new model.To address this, we propose N-GLARE (A Non-Generative, Latent Representation-Efficient LLM Safety Evaluator). N-GLARE operates entirely on the model’s latent representations, bypassing the need for full text generation. It characterizes hidden layer dynamics by analyzing the APT (Angular-Probabilistic Trajectory) of latent representations and introducing the JSS (Jensen-Shannon Separability) metric.Experiments on over 40 models and 20 red teaming strategies demonstrate that the JSS metric exhibits high consistency with Red Teaming safety rankings at less than 1% token and runtime cost.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1343ACL Anthology · ACL 2026

QiMeng-PRepair: Precise Code Repair via Edit-Aware Reward Optimization

Changxin Ke Institute of Computing Technology, Chinese Academy of Sciences, Rui Zhang Institute of Computing Technology, CAS, Jiaming Guo Institute of Computing Technology, Chinese Academy of Sciences, Yuanbo Wen Institute of Computing Technology, Chinese Academy of Sciences, Li Ding, Shuo Wang, Xuyuan Zhu, Xiong Peng 等 13 位作者

原文摘要

Large Language Models (LLMs) achieve strong program repair performance but often suffer from over-editing, where excessive modifications overwrite correct code and hinder bug localization. We systematically quantify its impact and introduce precise repair task, which maximizes reuse of correct code while fixing only buggy parts. Building on this insight, we propose PRepair, a framework that mitigates over-editing and improves repair accuracy. PRepair has two components: Self-Breaking, which generates diverse buggy programs via controlled bug injection and min–max sampling, and Self-Repairing, which trains models with Edit-Aware Group Relative Policy Optimization (EA-GRPO) using an edit-aware reward to encourage minimal yet correct edits. Experiments show that PRepair improves repair precision by up to 31.4% under \mathrm{fix}_1@1, a metric that jointly considers repair correctness and extent, and significantly increases decoding throughput when combined with speculative editing, demonstrating its potential for precise and practical code repair.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.2020ACL Anthology · ACL 2026

MIND: From Passive Mimicry to Active Reasoning through Capability-Aware Multi-Perspective CoT Distillation

Jin Cui Xi'an Jiaotong University, Jiaqi Guo Nankai University, Jiepeng Zhou, Ruixuan Yang, Jiayi Lu, Jiajun Xu Xi'an Jiaotong University, Jiangcheng Song, Boran Zhao 等 9 位作者

原文摘要

While Large Language Models (LLMs) have emerged with remarkable capabilities in complex tasks through Chain-of-Thought (CoT) reasoning, practical resource constraints have sparked interest in transferring these abilities to smaller models. However, achieving both domain performance and cross-domain generalization remains challenging. Existing approaches typically restrict students to following a single golden rationale and treat different reasoning paths independently. Due to distinct inductive biases and intrinsic preferences, alongside the student’s evolving capacity and reasoning preferences during training, a teacher’s “optimal” rationale could act as out-of-distribution noise. This misalignment leads to a degeneration of the student’s latent reasoning distribution, causing suboptimal performance. To bridge this gap, we propose MIND, a capability-adaptive framework that transitions distillation from passive mimicry to active cognitive construction. We synthesize diverse teacher perspectives through a novel “Teaching Assistant” network. By employing a novel Feedback-Driven Inertia Calibration mechanism, this network utilizes inertia-filtered training loss to align supervision with the student’s current adaptability, effectively enhancing performance while mitigating catastrophic forgetting. Extensive experiments demonstrate that MIND achieves state-of-the-art performance on both in-distribution and out-of-distribution benchmarks, and our sophisticated latent space analysis further confirms the mechanism of reasoning ability internalization.

cs.CL

Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) · 2026 · 正式发表