跳到正文
研踪PaperTrail

YOUR RESEARCH, IN SIGHT

让下一篇重要论文，不再错过。

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

部分采集异常
正在读取账号与研究记录…

全站更新概况

+0新增论文
+0其中 AI 相关
+1已有论文修订
+100新增 AI 导读

arXiv 最近完成扫描：2026-11-08 03:01 · 全站最近入库：2026-11-07 17:43各信源进度 →

统计口径与数据新鲜度

统计 2026-11-07 03:49 — 2026-11-08 03:49（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：2026-11-08 03:01最近发现新增版本：2026-11-07 17:43库内论文最新更新时间：2026-11-06 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。

17,100 已收录论文14,661 AI 相关4,056 有修订112 HF 社区发现

浏览论文库

研究日报 / 周报 →按信源浏览 →

时间范围跟随排序：版本更新、本站收录或首次发表。最新收录可能包含历史论文。部分信源异常，查看状态 →

找到 26 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-08 02:59（北京时间）
版本更新 · v2arXiv:2611.06748

EvoWorld: Towards Out-of-Sight World Evolution in Generative Video Models

Yequan Bie, Dengyang Jiang, Zhixuan Chen, Xiaozhuan Liang, Haoxuan Che, Quande Liu, Xi Chen, Hao Chen

原文摘要

Generative video models are increasingly used as visual world simulators, yet their primary strength lies in rendering plausible pixels rather than reasoning about how a world should evolve. Coherent world evolution requires more than visual fidelity: when an event occurs, a model must infer its consequences for the entities in the world and then render the resulting state. For example, after a wild beast appears, the model should not only depict the wild beast, but also infer whether a nearby person should flee, take cover, or fight back based on the realized scene. Existing diffusion-based video models lack an explicit mechanism for maintaining world state and reasoning over such event-conditioned consequences, often producing visually plausible but behaviorally static or causally inconsistent outputs. To make this reasoning gap concrete and measurable, we propose a new problem, Out-of-Sight Causal Evolution, which requires a model to render an event, reason about its consequences for the evolving world state, and reveal a causally plausible outcome as the world continues to evolve, even out of sight. As a baseline reference solution, we introduce EvoWorld, a brain-renderer collaboration framework in which a vision-language model serves as the world brain: it observes generated segments, maintains an evolving account of the world, and reasons about plausible state transitions, while a segment-autoregressive video renderer realizes them visually. We further introduce EvoWorld-Data and EvoWorld-Bench for training and evaluating event realization, causal reactions, and revisit consistency. Experiments show that EvoWorld substantially improves event-conditioned causal evolution over recent models while preserving visual quality and consistency, establishing a reference point for reasoning-guided, autonomous world evolution beyond the current field of view.

cs.CV · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.07192

Backbone-Informed Adaptation after EEG Pretraining: A Capacity-Controlled Study

Mingyang Jiang, Yamin Li, Fan Ma, Daniel Moyer, Hua Xu, Catie Chang

原文摘要

Electroencephalography (EEG) foundation models learn representations that combine information across channels and time. This structure motivates adapters that model the same relationships during downstream learning. However, it is not clear whether such a design outperforms other candidate adapters. Moreover, in order to arrive at a fair comparison between candidate adapters, we also need to consider their trainable parameters. Here, we ask whether a backbone-informed (native) adapter makes better use of a given capacity by comparing it against a token-wise multilayer perceptron control that transforms each feature vector independently. We approximately match their trainable parameter counts and keep the pretrained encoder fixed. Across two EEG foundation models and tasks covering emotion recognition, sleep staging, and EEG event classification, the native adapter shows no consistent advantage. For one backbone, differences between the designs are small relative to variation across runs. For the other, the token-wise control performs better on emotion recognition and EEG event classification. On sleep staging, the native adapter's advantage over the token-wise control is reduced when we also train the component that processes sequences of EEG segments. The relative performance of the two adapters depends on the backbone, task, and downstream training protocol. These findings motivate evaluating backbone-informed adapters against capacity-matched alternatives.

cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.06518

EmoReward: Self-Evolving Reward Calibration for Emotional Video Generation RL

Mingzhe Zheng, Yue Wu, Jing Tan, Chang Zou, Dengyang Jiang, Zeyue Tian, Xuanhua He, Ryan Xu 等 22 位作者

原文摘要

Group Relative Policy Optimization (GRPO) has advanced generative model post-training in verifiable domains such as code and mathematics, but evaluating video quality, especially emotional expression, often depends on subjective aesthetics and has no uniquely correct target. Existing approaches typically rely on fixed rewards for visual quality, motion quality, and text alignment that poorly capture emotional intent and can invite reward hacking, while RLHF is difficult to scale and subject to individual aesthetic differences that complicate the calibration of facial, body motion, and chromatic cues across narrative contexts. We introduce EmoReward, a new self-evolving training mechanism that distills evidence reliability from reward traces collected during an actual training run to adaptively guide multi-reward weight allocation in subsequent training by determining which emotional evidence to trust and how much to weight it. EmoReward first plans specific grounding cues from the emotional prompt to guide evaluation, then uses heterogeneous signal primitives to score the generated video during the act stage. Reflective Reward Gating calibrates their contributions through instance-level gates for phase, face confidence, conflict, and subtlety, then renormalizes weights over the active set to produce the sample's fused reward. RewardMemory then estimates each primitive's reliability from rollout scores and outcomes within each emotional regime in the finished training run, then fixes bounded multipliers that calibrate fusion weights in the next training run without updating reward models. Experiments on HunyuanVideo 1.5 show improvements across multiple RL algorithms and compare EmoReward with other agentic reward methods under SAGE-GRPO. Ablations and transfer experiments on SD 3.5 and FLUX further support adaptive reward calibration for visual generation.

cs.CV · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.06967

Call-Level Revision Contracts for LLM-Authored Workflows

Chenyang Jiang, Benyang Pan, Puqu Wu

原文摘要

Durable execution engines replay a recorded history and do not compare activity payloads. When a model repairs a workflow that failed partway through, it must either pay again for calls that succeeded or accept a recorded result whose payload changed. We define a revision contract for workflows whose orchestration code is written by a model. Each capability call carries a structural identity built from its callsite, its fanout position, and its occurrence count, and the identity stays stable under rewrites that leave the call in place, so committed identities form a frozen prefix. The frozen unit is the call expression, and the surrounding code stays writable. A static gate rejects a rewrite that would change a committed call before a container starts. A runtime digest stops a changed payload at a committed identity, and an occurrence ledger names a repeated paid payload. Under three assumptions, a candidate the gate accepts replays every committed call it invokes, executes only new work, and reports the identities it skips. No local identity is both relocation-invariant and free of wrong reuse, so callsite identity turns a relocation into a pre-spend stop. On media workflows, 22 gate-passed repairs replay 258 committed identities with no duplicate, and a live six-way fanout makes seven paid calls instead of twelve. A 12-task customer-support study applies the same identity, replay, and boundary checks on two models.

cs.SE

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.06061

TouchéWorld tech report : Towards Online Unified Camera--Hand Exploration in Generative Worlds

Jiaao Yu, Yanjian Zhang, Haoyang Jiang, Guillaume Wisniewski, Nadi Tomeh, Shijie Jia

原文摘要

Recent video world models allow users to explore generated environments through camera navigation. Continuous hand motion provides a fine-grained action channel for interacting with the people and objects encountered during exploration. We present TouchéWorld, a framework that integrates gesture-driven interaction into a camera-controlled video world model. Given an initial image and a scene-description prompt, the model generates video conditioned on temporally aligned camera and hand controls. We construct joint camera--hand training data from heterogeneous egocentric videos through annotation completion, consistency-based filtering, and temporal alignment. Building on a pretrained camera-controlled world model, we combine spatial hand-skeleton conditioning with numerical hand-motion features through parameter-efficient adaptation. Causal adaptation and few-step distillation support chunk-level online generation, allowing users to update their controls based on generated visual feedback. We further introduce TouchéBench to evaluate hand-motion following, camera--hand coordination, online interaction, and object response. Together, these components establish a framework for studying continuous exploration and hand-driven interaction within generative worlds.

cs.CV · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.05968

InterSim: Innovative Intersection Digital Twins for Human-in-the-Loop Driving Simulation

Luyang Jiang, Chuheng Wei, Yaguang Li, Shaoqi Wang, Yupeng Zhou, Hyunsoo Yun, Xishun Liao, Brandon Pitts 等 9 位作者

原文摘要

Innovative intersection designs can improve roadway safety and operational efficiency, but their uncommon geometries may create challenges for driver comprehension during early deployment. Evaluating these designs through full-scale field implementation can also be costly. This paper presents InterSim, a Unity game engine-based human-in-the-loop (HITL) driving simulator that constructs digital twins of site-specific innovative intersections in a virtual environment. InterSim links the real world and the simulation world through two closed cycles: a digital twin cycle that builds each intersection twin from agency design and traffic data and returns evidence from simulation to the agency, and an HITL cycle in which human drivers or automated driving algorithms operate the ego vehicle under identical scenario conditions. The research capabilities of InterSim are demonstrated through two case studies: a surrogate safety study compares post-encroachment time distributions between a Reduced Conflict Intersection and a standard intersection under different traffic volumes; an eco-driving study compares human drivers with a rule-based agent using vehicle-to-infrastructure information and reveals differences in emission and driving behavior. These studies demonstrate that InterSim provides a reusable digital twin platform for investigating innovative intersection designs from diverse perspectives. Demo video and open-source code can be accessed at https://purduedigitaltwin.github.io/InterSim/

cs.CE · cs.CY · cs.ET

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2611.04483

CogniLife: Continuous Cognitive Distillation for Deployed Multi-Expert Agent Marketplaces

Xuwei Tan, Jingchun Huang, Ming Cai, Shuangquan Lyu, Tiange Wang, Yinzhi Wang, Ronghua Shi, Jingwei Song 等 15 位作者

原文摘要

Multi-expert agent marketplaces -- where creators publish domain-specific AI experts for end users to consult -- are an emerging deployment pattern for language technology. A published expert can receive many requests yet little evidence about its creator, who remains active elsewhere: the cold-expert problem. We frame continuous cognitive distillation as converting attributed creator observations into revisable decision rules, communication patterns, and domain procedures. CogniLife proposes a cross-expert attribution design for this problem, with Creator Profiles for transferable creator patterns and Expert Journals for listing-specific knowledge; this design is not yet implemented or evaluated. A deployed platform partially realizes the surrounding lifecycle through creator-authorized authoring, capability-owned proposal filtering, and reversible consolidation of eligible owner-scoped memory. On ProfyBench-v1, an 85-task offline benchmark built from replay seeds and controlled internal tests, the Profy Full configuration as a whole achieves a mean overall score of 4.65/5 from the primary LLM judge, compared with 4.01 for Raw LLM and 4.28 for Expert Prompt, with the largest gains in completeness, specificity, and missing-information tasks. These configuration-level scores do not test cold-expert mitigation, cross-expert transfer, or longitudinal distillation.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.41275

TRACE: Target-State Risk-Aligned Conditional Execution for Embodied Tracking

Mingyang Lyu, Gongce Cao, Yiyang Jia, Lu Jia, Erliang Lin, Yuhao Zhang, Yinqian Sun, Pin Lyu 等 10 位作者

原文摘要

Embodied visual tracking requires a robot to follow a language-specified person under occlusion, look-alike distractors, and ambiguous instructions. We show that improving tracking persistence can simultaneously increase safety violations. Using a compact 0.6B vision-language-action (VLA) tracker with a frozen backbone, we update only the planner head with proximal policy optimization (PPO). PPO improves success and following rates across all three EVT-Bench settings, but increases the collision rate on distracted tracking because episodes that previously ended in target loss can instead end in contact. We introduce TRACE, an execution-time target-state risk gate that preserves the learned policy and intervenes selectively. A lightweight risk head processes 84 cues, including candidate boxes, identity margins, a learned range-and-bearing probe, and the policy's commands. When the calibrated risk exceeds a threshold, TRACE rescales the translational component of the next waypoint while leaving yaw unchanged. The executed action is identical to PPO's on more than 98% of steps. In paired evaluations on identical episodes, TRACE reduces the distracted-tracking collision rate by 3.96 percentage points, while further improving success and following rates. Matched slowdown controls without target-state timing do not reproduce this improvement. Concurrent evaluation reproduces a collision reduction of 3.77 points, compared with 0.29 points for identical-arm null pairs, indicating that the effect exceeds evaluation variability. These results show that target-state-aware execution can recover the safety cost of policy post-training with minimal intervention.

cs.RO

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.03203

MedCUA-Bench: A Screenshot-Only Benchmark for Clinical Computer-Use Agents

Jia Yu, Zilong Wang, Xinyang Jiang, Dongsheng Li, Shuo Wang

原文摘要

Computer-use agents could automate repetitive screen-based clinical work, but their reliability in medical graphical user interfaces remains largely unvalidated. Existing benchmarks focus on general web or desktop tasks and underrepresent medical software, which requires domain knowledge, exhibits markedly different UI design from mainstream applications, lacks public testing environments, and demands safety validation beyond task completion. We introduce MedCUA-Bench, an interactive benchmark for clinical computer-use agents. It covers 18 clinical scenarios across 10 medical domains, reconstructed from real product manuals and open-source medical systems to capture authentic clinical interfaces while avoiding licensing and privacy constraints. Each task ships with paired intent- and step-level goals to disentangle clinical reasoning from UI execution, and is evaluated by a deterministic checker over task completion and five clinical safety dimensions. Across 23 agents, the best closed-source model reaches 54.2% strict success, while all models remain below 9% on the real OpenEMR. Open-source agents average only 2.5%, with the best reaching 16.2%. MedCUA-Bench exposes the gap between current agents and reliable clinical software use, providing a reproducible testbed for future research.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.35013HF 社区推荐

Multiscale Policy Optimization for Long-Horizon LLM Agents

Tao Ren, Hui Yang, Jia Deng, Tingyang Chen, Qiyao Wang, Shuo Tang, Ziyang Zeng, Yuhua Jiang 等 19 位作者

原文摘要

Long-horizon LLM agents interact with their environments through complete assistant turns while generating each action token by token. We observe that critic values vary primarily across turns, motivating optimization anchored at turns with finer credit allocation among tokens. We propose Multiscale Policy Optimization (MPO), combining Multiscale Advantage Estimation (MAE) and turn-level importance sampling to align credit assignment and policy updates at the turn scale. Extensive experiments across search, tool-use, and automation tasks with 4B and 30B-A3B models demonstrate consistent gains over strong RL baselines.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2609.35203

NucShare: Lightweight Nuclei Instance Segmentation with Foundation-Model Accuracy

Wenyan Li, Xiangyang Jia, Wenlong Zhong, Xu Chen

原文摘要

Nuclei instance segmentation is fundamental to quantitative pathology. Yet high-accuracy methods mostly build on large vision transformers (ViTs) or heavy convolutional backbones, whose parameter count and compute cost hinder deployment in clinical and edge settings. We present NucShare, a lightweight model whose shared decoder handles nuclear pixel segmentation, horizontal-vertical distance, and classification through learnable transposed convolutions for upsampling and three lightweight heads, using 37% of independent decoders' parameters with no loss in accuracy. Output-level distillation then transfers knowledge from NucShare-T (UNI2-h) to NucShare-S (ConvNeXt V2-Tiny), reducing 690.5M parameters to 34.7M. On PanNuke they reach mPQ 0.524 and 0.519, above the prior best CFR-SAM-H at 0.511, and NucShare-S surpasses HoVer-Net and Mask R-CNN by 5.6 and 15.0 points; on zero-shot MoNuSeg it attains bPQ 0.668 against CFR-SAM-H's 0.662. Code: https://github.com/L-iMeng-L/Newidea

cs.CV

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.39497

Does Scaling Toward AGI Depend on What System-1 Lets System-2 Execute?

Qi Wang, Yun Qu, Lizhou Cai, Heming Zou, Haoang Chi, Yiqin Lv, Xiang Li, Haiyan Feng 等 13 位作者

原文摘要

Progress toward artificial general intelligence (AGI) has largely relied on scaling a resource-intensive executive, System-2, through larger models, more data and compute, longer reasoning, and more agents. We ask whether the next stage may come instead from co-designing it with System-1, a lightweight, semantically aware predictor that evaluates candidate interactions before execution, so that System-2 acts only on those worth their cost. Our answer is a qualified yes. We study \textit{conditional interaction}, a predict--gate--execute--update loop: System-1 estimates each candidate's utility, System-2 executes those admitted by an adaptive threshold, and the outcomes update System-1. The loop spans training, test-time reasoning, adaptive decision-making, and agentic tool use, with utility and cost defined by the application. Under ideal screening and equal execution costs, execution compute falls by a factor approaching $1/ρ_u$, where $ρ_u$ is the density of utility-relevant interactions (cost-weighted when costs differ); prediction errors, exploration, and distribution shift reduce the saving. We bound selection regret by predictor error and the utility margin. These results motivate our central hypothesis: $ρ_u$ is a measurable coordinate of scaling efficiency and makes informative compute a better description of scaling than total compute.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38882

Aligning Autoresearch: Co-Improvement between Agents and Humans

Tianjian Li, Chuanyang Jin, Seungone Kim, Tong Chen, John Hoffman, Yoram Bachrach amd Pedro Rodriguez, Ping Yu, Eryk Helenowski 等 13 位作者

原文摘要

Autoresearch agents can now attempt long-horizon AI research end-to-end, i.e., self-improvement. However, best results are still obtained by research agents interacting with human researchers, and improving AI's ability to perform this co-research will only reinforce this effect---a process termed co-improvement (Weston and Foerster, 2025). Further, aligning with human expectations is likely a safer path for research agents, especially given recent agent reward hacking phenomena (OpenAI, 2026b). In this work we address these issues in four ways: (i) we empirically study the role of humans in directing autoresearch agents, focusing on the domain of benchmark creation, and find humans still play a pivotal role particularly in the ideation phase; (ii) we also develop a benchmark, HackBench, that shows current autoresearch agents are not aligned and are unable to achieve hack localization sufficiently; (iii) we present a method to train agents for improved human interaction by using human simulators as a proxy during reinforcement learning (RL); and finally (iv) we develop a method to improve model text generation, and hence agent communication, by RL with rubrics aligned with human experts. Together, our training and evaluation recipes provide a path toward more capable and aligned (co-)research agents.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.14561

From Video to Vision: Learning Generalizable Humanoid Carrying from Human Motion

Jianming Lu, Ziluo Ding, Zhen Zhao, Shuyang Jin, Han Wen, Xinrun Xu, Zhiyuan Xu

原文摘要

Most existing approaches rely on video or motion-capture data for reinforcement learning (RL) training; however, all of these methods require an upper-level module to generate reference trajectories in real time. In object carrying tasks, the humanoid can only lift objects from designated locations along the reference trajectory, resulting in poor spatial generalization. Moreover, existing policies can only be evaluated inside a motion-capture studio. To address these limitations, we propose \textbf{MOVER}, a multi-stage, Teacher-Assistant-Student, RL training framework for humanoid object carrying. We first collect data from videos and employ GENMO and GMR to retarget the video data, combining this with data augmentation techniques to generalize the data across objects of varying shapes and sizes. In the \textbf{first stage}, building upon the DeepMimic \cite{peng2018deepmimic} motion-training framework, we train the humanoid to transport objects from specified locations along designated trajectories. In the \textbf{second stage}, we relax the tracking of the reference motion and adjust the reward so that the humanoid learns to transport objects from different positions. In the \textbf{third stage}, we apply a Dagger-PPO distillation method to distill away the proprioceptive position observations and reference joint position of the humanoid. This allows the deployed policy to operate without relying on global position information from a motion-capture system; instead, the humanoid can directly perceive the object's pose from its onboard RGB-D camera to complete the carrying task. We design a series of simulation experiments to compare the success rates across stages (Stages 1 vs. 2) and before versus after distillation. We further transfer the policy to a real humanoid, successfully transporting objects at varying positions in a motion-capture-free setting.

cs.RO

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.36756

Efficient Remote Sensing Image-Text Retrieval via Interaction-Free Recall and Text-Guided Reranking

Xi Chen, Xu Chen, Xiamgyang Jia, Wei Wang, Xu Zhang, Zhenyuan Shun

原文摘要

Remote sensing (RS) image-text retrieval plays a critical role in understanding massive RS imagery. However, the dense multi-object distribution and complex backgrounds in RS imagery make it difficult to simultaneously achieve fine-grained cross-modal alignment and efficient retrieval. Existing methods either rely on complex cross-modal interactions that lead to low retrieval efficiency, or depend on large-scale vision-language adaptation, which requires massive data and computational resources. To address these issues, we develop a fast-then-fine (FTF) retrieval framework centered on interaction-free multi-granular recall and parameter-free text-guided cross-modal reranking with the Balanced Text-guided Interaction Block (BTIB). Specifically, interaction-free multi-granular recall pre-computes gallery embeddings and matches them with the corresponding query embedding without full-gallery token-level text-image interaction, while text-guided cross-modal reranking with BTIB refines local semantic correspondence within the top-K candidate set. Furthermore, an inter- and intra-modal loss is designed to jointly optimize cross-modal alignment across multi-granular representations. Extensive experiments on public benchmarks demonstrate that the FTF achieves competitive retrieval accuracy while significantly improving retrieval efficiency compared with existing methods.

cs.CV

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.34839

Behavioural Quality Is Not Learning Utility: Objective Mismatch in Human Behaviour Models for Robot Learning

Haoyang Jiang, Dana Kulic, Michael Burke, Pamela Carreno-Medrano

原文摘要

Human behaviour models are increasingly used to provide simulated human interactions for training robots via reinforcement learning (RL). Existing approaches primarily evaluate these models according to how closely their generated behaviours resemble real human data, while their usefulness as training partners is less commonly assessed. This creates a potential objective mismatch: a model that performs well according to behavioural evaluation metrics may not necessarily provide useful training experiences for downstream robot policy learning. Across two human-robot interaction (HRI) domains, highway driving and object handover, we evaluate different types of human behaviour models using common generative-model evaluation metrics. We consider multiple human trajectory representations and compare the resulting evaluations with the downstream performance of RL policies trained using the corresponding models. We find that existing behavioural evaluation metrics do not consistently predict downstream policy performance across domains and trajectory representations, while their evaluations are sensitive to the choice of trajectory representation. We also show that even when the true testing behaviour distribution is known, training against a different distribution may yield better downstream performance than training directly against the true distribution.

cs.RO

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.26030HF 社区推荐

Truncate or Not? A Scaling Study of Looped Transformers with Truncated Backpropagation

Xuyuan Xiong, Siyu Chen, Zihan Shao, Yue Su, Weiyang Jin, Ping Luo, Zhuoran Yang, Xihui Liu

原文摘要

Looped transformers reuse a shared block of layers to increase computational depth without increasing the number of trainable parameters. They are trained with backpropagation through time (BPTT). To save training compute, BPTT is often truncated to the last few loops, but truncation yields biased gradients that may weaken the benefit of looping. In this work, we study the value of truncation in looped transformer training in a systematic way: we jointly vary model size, training tokens, and forward and backward loop counts across nearly 600 training runs and fit a unified scaling law that captures the effects of looping and truncation through effective parameters. Despite the training-compute savings of truncation, our fitted law favors full BPTT when forward and backward loop counts are jointly optimized. In matched-compute runs, full BPTT through fewer loops consistently outperforms truncated settings of the same compute. The law further shows that at fixed model size, the compute-optimal loop count and training-token budget both grow with compute. When model size is also free, the law identifies two fully trained loops as the best looped setting. Experimentally, this setting achieves lower validation loss than the compute-optimal non-looped model with about 27% fewer non-embedding parameters and only 8% more training compute.

cs.LG · stat.ML

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2606.23525Apple ML Research

自压缩语言模型智能体

Self-Compacting Language Model Agents

Tianjian Li, Jingyu Zhang, William Jurayj, Xi Wang, Chuanyang Jin, Mehrdad Farajtabar, Eric Nalisnick, Daniel Khashabi

AI 速读 基于摘要

提出SelfCompact框架，让语言模型智能体在推理时自主决定何时及如何压缩上下文，以降低token成本并提升性能。

cs.CL

arXiv ↗PDF ↗
查看原文摘要

Long agent traces composed of chains of thought and tool calls accumulate stale content that anchor subsequent generations, and eventually outgrow the context window. Existing scaffolds mitigate it with fixed-interval compaction triggered at a token threshold. Such triggers pay no heed to trajectory structure, risking discard of partial results mid-derivation or mid-search. We propose SelfCompact, a scaffold that allows the model itself to decide when and how to compact. Specifically, it pairs two inference-time elements: (i) a compaction tool the model invokes to summarize the accumulated context, and (ii) a lightweight rubric specifying when to fire (a sub-task has resolved, or the trajectory is converging) and when to suppress (mid-derivation, or when stuck). Both are needed. The tool alone is unevenly used across open-weight models, often invoked at unhelpful moments or not at all; the rubric alone cannot act. Together, they elicit effective adaptive compaction without any fine-tuning or external supervision. We present empirical results on six benchmarks (competitive math and agentic search) and seven models. Our results show that SelfCompact matches or exceeds fixed-interval summarization at a fraction of the token cost, improving over a no-summarization baseline by up to 18.1 points on math and 5-9 points on agentic search at 30-70% lower per-question cost. Our results expose a meta-cognitive gap: although unprompted models cannot reliably tell when their own context is rotting, a lightweight rubric closes this gap, reframing when to compact as a capability that scaffolds can supply without training.

首次观测 · 记录 #1ACL Anthology · ACL 2026:acl:2026.acl-long.1626ACL Anthology · ACL 2026

Specializing Large Models for Oracle Bone Script Interpretation via Component-Grounded Multimodal Knowledge Augmentation

Jianing Zhang, Runan Li, Honglin Pang, Ding Xia The University of Tokyo, Zhou Zhu, Qian Zhang, Chuntao Li Jilin University, Xi Yang Jilin University

原文摘要

Deciphering ancient Chinese Oracle Bone Script (OBS) is a challenging task that offers insights into the beliefs, systems, and culture of the ancient era. Existing approaches treat decipherment as a closed-set image recognition problem, which fails to bridge the “interpretation gap”: while individual characters are often unique and rare, they are composed of a limited set of recurring, pictographic components that carry transferable semantic meanings. To leverage this structural logic, we propose an agent-driven Vision-Language Model (VLM) framework that integrates a VLM for precise visual grounding with an LLM-based agent to automate a reasoning chain of component identification, graph-based knowledge retrieval, and relationship inference for linguistically accurate interpretation. To support this, we also introduce OB-Radix, an expert-annotated dataset providing structural and semantic data absent from prior corpora, comprising 1,022 character images (934 unique characters) and 1,853 fine-grained component images across 478 distinct components with verified explanations. By evaluating our system across three benchmarks of different tasks, we demonstrate that our framework yields more detailed and precise decipherments compared to baseline methods.

cs.CL · cs.CV

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.2068ACL Anthology · ACL 2026

Benchmarking Fine-Grained Error Detection in Multimodal Reasoning

Chi-Min Chan, Han Zhu, Chunyang Jiang, Jiaming Ji, Juntao Dai, Wei Xue Hong Kong University of Science and Technology, Sirui Han The Hong Kong University of Science and Technology, Yike Guo Hong Kong University of Science and Technology and Imperial College London

原文摘要

Multimodal Process Reward Models (MPRMs) have emerged as a pivotal framework for enhancing the reasoning capabilities of Multimodal Large Language Models (MLLMs). However, the research community currently lacks a dedicated benchmark to rigorously assess the error discernment capabilities of these models.To address this gap, we introduce PRMBench-V, a novel benchmark specifically designed to evaluate MPRMs’ proficiency in detecting erroneous reasoning steps across diverse error categories. Leveraging a semi-automated annotation pipeline augmented with human verification, we construct a comprehensive dataset comprising 907 unique queries, each annotated with nine distinct error types, resulting in 8,163 test cases with fine-grained step-level error labels.Through extensive experiments involving over 15 open- and closed-source models, we uncover several key findings: (1) even the strongest existing MPRMs achieve only \textasciitilde30\% accuracy in error identification; (2) while partial error detection achieves moderate precision and recall (\textasciitilde60\%), overall accuracy remains low (\textasciitilde20\%); and (3) benchmark scores exhibit a strong correlation with downstream task performance gains (r=0.86). Furthermore, we demonstrate that PRMBench-V can inform the development of more robust MPRMs: by introducing the Bayesian Rater Reliability Process Reward Model (BR2-PRM), we achieve up to a 4.8% performance improvement through test-time scaling.We believe that PRMBench-V will serve as a valuable resource for advancing MPRM research, enabling more rigorous evaluation and fostering the development of models with fine-grained multimodal reasoning capabilities.

cs.CL · cs.CV

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.237ACL Anthology · ACL 2026

Miner: Mining Intrinsic Mastery for Data-Efficient RL in Large Reasoning Models

Shuyang Jiang, Yuhao Wang, Ya Zhang Shanghai Jiao Tong University, Yanfeng Wang Shanghai Jiao Tong University, Yu Wang Shanghai Jiao Tong University

原文摘要

Current critic-free RL methods for large reasoning models suffer from severe inefficiency when training on positive homogeneous prompts (where all rollouts are correct), resulting in waste of rollouts due to zero advantage estimates. We introduce a radically simple yet powerful solution to Mine intrinsic mastery (Miner), that repurposes the policy’s intrinsic uncertainty as a self-supervised reward signal, with no external supervision, auxiliary models, or additional inference cost. Our method pioneers two key innovations: (1) a token-level focal credit assignment mechanism that dynamically amplifies gradients on critical uncertain tokens while suppressing overconfident ones, and (2) adaptive advantage calibration to seamlessly integrate intrinsic and verifiable rewards. Evaluated across six reasoning benchmarks on Qwen3-4B and Qwen3-8B base models, Miner achieves state-of-the-art performance among the other four algorithms, yielding up to 4.58 absolute gains in Pass@1 and 6.66 gains in Pass@K compared to GRPO. Comparison with other methods targeted at exploration enhancement further discloses the superiority of the two newly proposed innovations. This demonstrates that latent uncertainty exploitation is both necessary and sufficient for efficient and scalable RL training of reasoning models. Code is available at https://github.com/pixas/Miner.

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.276ACL Anthology · ACL 2026

OCP: Outlier-Centric Probing for Dynamic Structured Pruning of LLMs

Yang Ji, Ying Sun Hong Kong University of Science and Technology (Guangzhou)

原文摘要

Structured pruning offers a hardware-friendly approach for efficient LLM inference. Early static methods determine fixed subnetworks through offline calibration, suffering from performance degradation and calibration sensitivity. Recent methods explore input-adaptive pruning by selecting a subset of tokens as probes to estimate hidden activations for online pruning decisions.However, existing probe selection strategies fail to identify outlier-triggering tokens, and uniform layer-wise sparsity misaligns with heterogeneous outlier distributions, leading to critical channels being incorrectly pruned. Therefore, we propose OCP (Outlier-Centric Probing for structured pruning), a principled framework that prioritizes capturing outlier-triggering tokens rather than reconstructing full hidden distributions. Specifically, OCP includes three key components: (1) sensitivity-weighted probing for FFN layers that identifies outlier patterns via precomputed weight aggregations, (2) attention-accumulated probing that leverages preceding attention matrices to identify salient tokens, and (3) online adaptive sparsity allocation that dynamically adjusts layer-wise pruning based on history-guided outlier statistics. Extensive experiments on LLaMA2, LLaMA3, and OPT demonstrate that OCP consistently outperforms state-of-the-art methods across benchmarks, achieving up to 25% perplexity reduction at 1.6\times speedup.

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.367ACL Anthology · ACL 2026

GAMBIT: A Gamified Jailbreak Framework for Multimodal Large Language Models

Xiangdong Hu, Yangyang Jiang, Qin Hu Shandong University, Xiaojun Jia Nanyang Technological University

原文摘要

Multimodal Large Language Models (MLLMs) have become widely deployed, yet their safety alignment remains fragile under adversarial inputs. Previous work has shown that increasing inference steps can disrupt safety mechanisms and lead MLLMs to generate attacker-desired harmful content. However, most existing attacks focus on increasing the complexity of the modified visual task itself and do not explicitly leverage the model’s own reasoning incentives. This leads to them underperforming on reasoning models (Models with Chain-of-Thoughts) compared to non-reasoning ones (Models without Chain-of-Thoughts). If a model can think like a human, can we influence its cognitive-stage decisions so that it proactively completes a jailbreak? To validate this idea, we propose GAMBIT (Gamified Adversarial Multimodal Breakout via Instructional Traps), a novel multimodal jailbreak framework that decomposes and reassembles harmful visual semantics, then constructs a gamified scene that drives the model to explore, reconstruct intent, and answer as part of winning the game. The resulting structured reasoning chain increases task complexity in both vision and text, positioning the model as a participant whose goal pursuit reduces safety attention and induces it to answer the reconstructed malicious query. Extensive experiments on popular reasoning and non-reasoning MLLMs demonstrate that GAMBIT achieves high Attack Success Rates (ASR), reaching 92.13% on Gemini 2.5 Flash, 91.20% on QvQ-MAX, and 85.87% on GPT-4o, significantly outperforming baselines. Warning: This paper contains unsafe and offensive examples.

cs.CL · cs.CV

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.454ACL Anthology · ACL 2026

Think in Latent Thoughts: A New Paradigm for Gloss-Free Sign Language Translation

Yiyang Jiang, Li Zhang, Xiao-Yong Wei, Li Qing

原文摘要

Many SLT systems quietly assume that brief chunks of signing map directly to spoken-language words. That assumption breaks down because signers often create meaning on the fly using context, space, and movement. We revisit SLT and argue that it is mainly a cross-modal reasoning task, not just a straightforward video-to-text conversion. We thus introduce a reasoning-driven SLT framework that uses an ordered sequence of latent thoughts as an explicit middle layer between the video and the generated text. These latent thoughts gradually extract and organize meaning over time. On top of this, we use a plan-then-ground decoding method: the model first decides what it wants to say, and then looks back at the video to find the evidence. This separation improves coherence and faithfulness. We also built and released a new large-scale gloss-free SLT dataset with stronger context dependencies and more realistic meanings. Experiments across several benchmarks show consistent gains over existing gloss-free methods. Our code and data are available at https://github.com/fletcherjiang/SignThought.

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.459ACL Anthology · ACL 2026

VoxMind: An End-to-End Agentic Spoken Dialogue System

Tianle Liang, Yifu Chen Zhejiang University, Shengpeng Ji HY LLM Team, Yijun Chen, Zhiyang Jia, Jingyu Lu, Fan Zhuo, Xueyi Pu 等 10 位作者

原文摘要

Recent end-to-end spoken dialogue models enable natural interaction. However, as user demands become increasingly complex, models that rely solely on conversational abilities often struggle to cope. Incorporating agentic capabilities is therefore essential: by enabling tool use, these models can extend their knowledge boundaries and better solve real-world tasks. Yet, existing research has largely concentrated on core perception and generation, with comparatively limited exploration of such tool-augmented extensions. To bridge this gap, we present VoxMind, an integrated framework designed to equip end-to-end spoken dialogue models with comprehensive agentic abilities. Leveraging our curated 470-hour AgentChat dataset, we incorporate a “Think-before-Speak” mechanism, enabling the model to internalize structured reasoning as a critical prerequisite for planning and response generation. Furthermore, to mitigate latency bottlenecks caused by large-scale tool integration, we propose a Multi-Agent Dynamic Tool Management architecture. By asynchronously delegating retrieval tasks to an auxiliary agent aligned with the main model’s reasoning trajectory, this system effectively decouples inference latency from toolset size. Experimental results confirm that VoxMind achieves significant improvements in agent performance: compared with strong baselines, the task completion rate increases from 34.88% to 74.57%, outperforming Gemini-2.5-Pro on spoken agent tasks while preserving general conversational quality. The source code and associated data are publicly available at https://github.com/MM-Speech/VoxMind.

cs.CL · cs.AI · cs.MA

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.770ACL Anthology · ACL 2026

MultiFinBen: Benchmarking Large Language Models for Multilingual and Multimodal Financial Application

Xueqing Peng Yale University, Lingfei Qian Yale University, Yan Wang Yale University, Ruoyu Xiang, Yueru He Columbia University, Yang Ren Yale University, Mingyang Jiang, Vincent Jim Zhang 等 47 位作者

原文摘要

Real-world financial analysis involves information across multiple languages and modalities, from reports and news to scanned filings and meeting recordings. Yet most existing evaluations of LLMs in finance remain text-only, monolingual, and largely saturated by current models. To bridge these gaps, we present MultiFinBen, the first expert-annotated multilingual (five languages) and multimodal (text, vision, audio) benchmark for evaluating LLMs in realistic financial contexts. MultiFinBen introduces two new task families: multilingual financial reasoning, which tests cross-lingual evidence integration from filings and news, and financial OCR, which extracts structured text from scanned documents containing tables and charts. Rather than aggregating all available datasets, we apply a structured, difficulty-aware selection based on advanced model performance, ensuring balanced challenge and removing redundant tasks. Evaluating 21 leading LLMs shows that even frontier multimodal models like GPT-4o achieve only 46.01% overall, stronger on vision and audio but dropping sharply in multilingual settings. These findings expose persistent limitations in multilingual, multimodal, and expert-level financial reasoning. All datasets, evaluation scripts, and leaderboards are publicly released.

cs.CL · cs.CV

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