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

+1199新增论文
+950其中 AI 相关
+132已有论文修订
+97新增 AI 导读

arXiv 最近完成扫描:2026-11-20 08:48 · 全站最近入库:2026-11-20 16:53各信源进度 →

统计口径与数据新鲜度

统计 2026-11-19 18:11 — 2026-11-20 18:11(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:2026-11-20 08:48最近发现新增版本:2026-11-20 16:53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。

28,609 已收录论文23,738 AI 相关8,785 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 78 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-20 08:46(北京时间)
新论文 · v1arXiv:2611.26366

World Models Favor Learned Priors over Visual Evidence

Han Xiao, Zhiyong Bao, Xiaotong Liu, Yanhao Chen, Qingqiang Wu

原文摘要

World models are increasingly used to simulate physical interactions, but training a separate model for each environment limits scalability. Cross-environment prediction requires identifying the current physical rule, which appearance alone may not reveal. We therefore study visual calibration: using object motion and interactions in a short video as direct evidence to infer the current rule and predict motion under that rule from a different initial state. However, differences in data availability across physical conditions can yield a long-tailed training distribution, biasing models toward common rules even when evidence supports rare ones. To study this challenge, we introduce TailWorld, which evaluates the transfer of demonstrated rules across initial states while independently controlling training imbalance and calibration informativeness. The benchmark compares predictions with two reference futures from the same prediction initial state, one following the demonstrated rule and the other a common rule, to measure prediction bias toward common dynamics. We further propose Rule Inference for Transfer (RIFT), which learns how different rules affect object motion, infers the current rule from calibration interactions, and applies it to prediction from a new initial state. RIFT explicitly controls the influence of training frequency on rule inference, allowing sufficiently informative observational evidence to support rare rules. Experiments on TailWorld show that RIFT improves prediction accuracy and physical consistency across varying training imbalances and distribution shifts, with gains in predicting rare dynamics.

cs.AI

版本更新 · v6arXiv:2603.28584

ORSIFlow: Saliency-Guided Rectified Flow for Optical Remote Sensing Salient Object Detection

Haojing Chen, Zhihang Liu, Yutong Li, Tao Tan, Haoyu Bian, Qiuju Ma

原文摘要

Optical Remote Sensing Image Salient Object Detection (ORSI-SOD) remains challenging due to complex backgrounds, low contrast, irregular object shapes, and large variations in object scale. Existing discriminative methods directly regress saliency maps, while recent diffusion-based generative approaches suffer from stochastic sampling and high computational cost. In this paper, we propose ORSIFlow, a saliency-guided rectified flow framework that reformulates ORSI-SOD as a deterministic latent flow generation problem. ORSIFlow performs saliency mask generation in a compact latent space constructed by a frozen variational autoencoder, enabling efficient inference with only a few steps. To enhance saliency awareness, we design a Salient Feature Discriminator for global semantic discrimination and a Salient Feature Calibrator for precise boundary refinement. Extensive experiments on multiple public benchmarks show that ORSIFlow achieves state-of-the-art performance with significantly improved efficiency.

cs.CV

新论文 · v1arXiv:2611.25478

TAP: Tapping the Discriminative Potential of Frozen Representations for Class-Incremental Learning

Zihao Wang, Yu-Wei Zhan, Xiao-Qian Liu, Zhen-Duo Chen, Weitong Liu, Xin Luo, Xin-Shun Xu

原文摘要

Pre-trained models provide transferable representations for class-incremental learning (CIL). However, sequentially adapting these representations introduces optimization costs and may disrupt old-class feature structure. This motivates us to explore whether the discriminative potential of these representations can be further exploited by continually accumulating class information while keeping the encoders frozen. To this end, we propose TAP, a new and novel framework for tapping the discriminative potential of frozen representations in class-incremental learning. It accumulates visual and semantic class information in the frozen CLIP embedding space with only lightweight optimization. TAP comprises three modules: class-conditional visual distribution modeling, class-adaptive semantic calibration, and visual-semantic posterior fusion. The first module retains class means and within-class scatters, combining class-specific and shared covariances to stabilize distribution estimation. The second module learns class-specific weights over text templates from current-task images, adapting text representations to downstream classes. The third module combines the visual and semantic posteriors through a weighted product of experts for prediction over all seen classes without retaining images from previous tasks. Across nine datasets under two incremental protocols, TAP achieves state-of-the-art performance, attaining the highest average incremental and final-stage accuracies among the compared methods. The code is provided as an ancillary file.

cs.CV

新论文 · v1arXiv:2611.25469

A Wolf in Sheep's Pixels: Visual Camouflage Attacks via HDR Gain Maps

Qi Shen, Yujia Gong, Yue Zhao, Tong Liu

原文摘要

An image can pass automated safety checks while showing harmful content to its human viewer. We introduce DRJanus, a visual camouflage attack that exploits the gap between standard dynamic range (SDR) processing and high dynamic range (HDR) display. Automated systems often read only an image's SDR base, whereas HDR viewers reconstruct its appearance using an auxiliary gain map. DRJanus inverts this reconstruction to embed two independently chosen color images in a single JPEG, without querying or optimizing against the target model. It preserves the chosen SDR base exactly and reconstructs HDR targets at an average PSNR of 43.03 dB across 400 images. On a benchmark of 100 sexual-content images, DRJanus evades detection by all six tested commercial moderation services and nine vision-language models. Across five agents, each evaluated on 120 task cases per threat setting, attack success ranges from 31.67% to 100% for moderation evasion and from 95% to 100% for task manipulation. We also explore cross-view consistency checks as a mitigation. These findings expose a gap in visual trust: securing what a machine reads does not secure what a user sees. A demo is available at https://qshen0629.github.io/DRJanus/

cs.CV · cs.AI

新论文 · v1arXiv:2611.25419

DiLE: Discrimination-Aware Latent-Space Learning for Universal Multimodal Embeddings

Ben Kang, Junda Wang, Yuxuan Yuan, Hang Li, Yuhan Liu, Xin Chen, Wanting Geng, Tong Lin 等 12 位作者

原文摘要

Multimodal large language models have greatly improved input understanding in universal multimodal embedding models. However, this understanding does not ensure that their embeddings preserve the evidence needed to distinguish relevant items from confusable alternatives. To preserve essential discriminative evidence in embeddings, this paper proposes DiLE, a discrimination-aware latent-space learning framework that combines task guidance with on-policy retrieval feedback. Specifically, a few learnable queries are used to encode task-relevant information in a compact latent space. Reasoning instructions and auxiliary answer-generation supervision shape these representations to support both task understanding and retrieval discrimination. To address task-specific discriminative needs, we design shared-private expert adaptation that integrates cross-task knowledge sharing with task-specific representation learning. Building on this, the on-policy hard-negative distillation is introduced: the student retrieves confusable candidates, and a reverse KL divergence objective aligns its embedding-based similarity distribution with the teacher's fine-grained relevance distribution over these candidates. Candidates are periodically refreshed using updated student snapshots, tying teacher supervision to the student's evolving retrieval behavior and confusions. Across four model configurations that span two backbone generations, DiLE achieves state-of-the-art performance on MMEB-v2, UVRB, and MRMR, demonstrating its effectiveness across modalities and tasks.

cs.CV

新论文 · v1arXiv:2611.24785

Read Together, Answer Apart: How Earlier Questions Help Vision-Language Models Look

Xinyi Guo, Kaile Huang, Xitong Ling, Mingyi He, Chende Zheng, Jianyu Xu, Chao He, Dongqing Song

原文摘要

Vision-language models (VLMs) are often asked several questions about the same image in a single prompt (co-asking), yet existing evaluations report conflicting accuracy gains and hallucinations at the prompt level without tracking how individual answers change. By pairing each question's single-question and co-asked answers across 22 open-weight VLMs (1B-35B), we find that erratic net shifts reflect two opposing effects: on Densely Captioned Images (DCI), every model repairs 15.4%-29.8% of its single-question errors under co-asking, while the share of correct answers turned wrong falls from 30.8% on weaker models to 4.7% on stronger ones. Across all 14 models supporting custom attention masks, we show through attention readouts and layer-window interventions that earlier same-image questions read visual information into their question tokens in the early-to-middle layers during prefill, helping later questions concentrate image attention onto their queried targets. This visual help accumulates across the scene even across non-overlapping regions, whereas blocking earlier questions from attending to the image lowers both target localization and accuracy in all 14 models. While cross-question help is established during shared question reading, part of the co-asking damage enters later during answer decoding when the decoder attends to other question stems. We therefore introduce Read Together, Answer Apart, a training-free decoding mask that keeps shared prefill intact and isolates each answer from other stems, improving co-asking accuracy on 12 of 14 models on DCI (up to +4.71 percentage points, closing 70% of the pooled gap to asking alone) and on 10, 8, and 5 of the 14 models on WebUI, FinTabNet, and GQA (up to +11.4, +5.3, and +1.1 points).

cs.CV

新论文 · v1arXiv:2611.24381

How to Turn an Attention Neural Network into an Interpretable Hard-Logic Predicate Circuit: Truth-Table Predicate Extraction and State-Machine Merging under Ship-of-Theseus Replacement

Yuchen Li, Zongqi Fan, Xintong Li, Nguyen H. Tran, Ken-Tye Yong

原文摘要

We convert a trained attention network as a whole, attention included, into a hard-logic predicate circuit that runs in its place. Every unit of the circuit is a 0/1 predicate on tokens, and every map between predicates is a minimised truth table. Short tables, such as the product of a query code and a key code, are minimised exactly, and wide feature-to-feature transitions become implication rules by noise-tolerant minimisation. Blocks are replaced one at a time from the output down, in the manner of the Ship of Theseus, and after each replacement the remaining soft part re-adapts to the frozen circuit. Attention selection takes two forms, a row-wise comparison of integer scores (order selection) and a learned rule (rule selection). Both are fitted on the ordinary training set alone, and only on the examples that the teacher classifies correctly, so that the circuit imitates the teacher where the teacher is right. On TweetEval-emotion, the complete circuit of DistilGPT2 reaches a test gold accuracy of 66.5 $\pm$ 1.3% with order selection and 57.4 $\pm$ 2.5% with rule selection over three seeds, against 74.7% for the teacher and 64.0% for a bag of word pieces fitted on the same tweets, and that of BERT-base reaches 70.0 $\pm$ 0.5% with order selection and 63.7 $\pm$ 1.2% with rule selection against 77.8%. The circuit needs no multiplier and no weight traffic, and its exported hardware description reproduces it bit for bit. By a synthesis-based estimate it classifies one tweet in about 1-2 $μ$s (DistilGPT2) and 3-5 $μ$s (BERT) on an FPGA card and in 5-9 $μ$s on a palm-sized board whose module draws 7.5-15 W, two to three orders of magnitude faster than the network on a GPU at batch 1 (measured, 0.9-2.6 ms for DistilGPT2 including an optimised CUDA-graph baseline), at a cost of 8.5 (DistilGPT2) and 8.6 (BERT) points of gold accuracy for the synthesised seed-0 circuits.

cs.LG · cs.AR

新论文 · v1arXiv:2611.24193

WM^MLLM: Turning MLLMs into World Models

Baiang Li, Enze Xie, Jincheng Yu, Jin Wang, Haozhe Liu, Yitong Li, Juanxi Tian, Felix Heide

原文摘要

Video world models keep the world in the context of a generator, so a revisit is synthesized again and earlier edits can drift or disappear. We present WM^MLLM, the first interactive world model built on multimodal large language models (MLLMs), for persistent, editable world generation from a single image. WM^MLLM keeps the world state outside every generator, as a panorama with a revision history, and lets the MLLMs write it: they turn each request into operations on this state, select where a local edit is written, and check every candidate before it becomes the next state. Image generators only propose views, and video only presents them. Users can walk to new places and return, edit objects through text or a box, place subjects, and change lighting, weather or season. By construction, a revisit at the same revision reproduces the earlier view exactly, a local edit leaves every panorama pixel outside its region unchanged, and each request starts from the result of the previous one. Against six world model baselines on MBench-A, WM^MLLM has the highest mean on all four environment-recovery metrics (Epipolar 91.5 versus 0.1 to 74.0) and completes the most camera round trips (85.3% versus at most 50.9%). It has the best value in all 15 CompBench columns against 15 baselines. On 112 local edits judged blind, selecting the region to write after generation raises completion from 34.8% to 83.0%. On 258 WBench and MBench-A items, all 46 revisits recover their earlier views, and all 97 local edits that pass the check leave panorama pixels outside their region unchanged.

cs.CV

新论文 · v1arXiv:2611.24006

MARS Technical Report

Tong Liu, Chaoqun Hou, Longxiang Xiong, Junhong Lou, Jiaqi Zheng, Qixin Li, Zixin Zhao, Yong Duan 等 17 位作者

原文摘要

Modern industrial recommender systems use multi-stage cascades--retrieval, pre-ranking, ranking, and re-ranking--to narrow millions of candidates to a personalized list. While each stage deploys sophisticated models, pipeline-level decisions on channel activation, quotas, model selection, and score fusion remain engineer-tuned and user-agnostic within each scenario. This report presents MARS (Multi-stage Agentic Recommendation System), extending personalization from individual-stage predictions to full pipeline orchestration. A near-line agent built on fine-tuned Qwen3-0.6B generates per-user configurations--a seven-dimensional action covering pipeline mode, retrieval quotas, model selection, fusion weights, and reranker settings--from a constrained action space. Configurations are validated, cached, and consumed asynchronously; the cascade never waits for LLM inference. Schema checks, fallbacks, and a kill switch ensure serving safety. The policy is trained in three stages: Confidence-aware SFT distills schema-valid generation from a teacher model; Format-Anchored Preference Optimization (FAPO) learns action preferences via ActionMask with posterior weighting; Group Relative Policy Optimization (GRPO) refines the policy through Sandbox replay with a frozen reward model, without exposing exploratory actions to live traffic. Two-week A/B tests across five Taobao scenarios with over ten million daily users show positive gains on all key metrics. The largest reach +5.22% V2P and +7.34% GMV, with positive CTR lifts on both click surfaces. About 20% of requests use a lightweight Minimal mode, and deployment shows 12 ms lower mean serving latency versus production.

cs.IR

新论文 · v1arXiv:2611.23606

VIBRA: Backward-Compatible Cross-Modal Retrieval with Consensus Barycenters

Shubing Yang, Xutong Liu, Dongfang Zhao

原文摘要

Backward-compatible retrieval enables a new encoder to query the gallery that an old model has already encoded, so that a cross-modal retrieval system can be upgraded without re-encoding billions of items. Existing compatible methods either fit the old query and stay near the accuracy of the old model, or run the old and the new encoder for every query. We propose VIBRA, which matches relevance on the old gallery, corrects the errors of a single model with the barycenter of several views, which is itself a query in the old space, and trains the new encoder to inherit this consensus, so that models absent at deployment still shape every query. We show that the barycenter gains margin over the average of its views exactly when their hardest irrelevant items differ, and we characterize the share of this gain that one encoder can reach. On five benchmarks with five upgrades among six CLIP, MetaCLIP and SigLIP2 backbones, VIBRA is the best compatible method in all 50 settings, improving Recall@1 by up to 5.6 points over the strongest baseline and by up to 18.0 points over the old model, while adding less than 1 ms to the latency of the new encoder.

cs.CV · cs.IR

版本更新 · v3arXiv:2602.07796

Thinking Is Not Telling: Information Disclosure in User-Service LLM Agents

Jiatong Li, Changdae Oh, Hyeong Kyu Choi, Jindong Wang, Sharon Li

原文摘要

User-engaged LLM agents increasingly operate in service scenarios where task success depends on coordination between the agent, the user, and a stateful environment. In such interactions, the agent often has access to rich information that the user does not observe, such as task policies, tool results, and environment states. This makes agent-user communication a central component of task completion. In this work, we study a communication-level failure mode of user-service agents: internal reasoning is counterproductive for the interactive information disclosure. We introduce information disclosure as a measurable communication mechanism, covering state corrections, available options, constraints, consequences, and proactively retrieved environment information. Motivated by recent advances in LLM reasoning, we evaluate and analyse information disclosure under a controlled setting of agents equipped with a "think" tool, which allows agents to reason internally before responding or acting. Across five models and three user-service scenarios, we first show that enforced thinking has unstable utility. We then analyse response scaling under different think efforts and find that more thinking does not reliably translate into richer information disclosure. Using turn-level counterfactual intervention, we show that injecting golden decision-relevant information significantly intervenes agents to correct action trajectories, indicating that missing disclosure can causally affect the agent's immediate downstream behaviour. We further find that failed trajectories most often lack state corrections and constraints, and that a lightweight disclosure-aware prompt effectively mitigates degradation. These results suggest that the key challenge is not simply making agents think more, but ensuring that reasoning and information are externalised as timely information disclosure.

cs.CL

新论文 · v1arXiv:2611.22971

SEAM: Computer-Use Agents Know the Harm Between Screens, and Commit It Anyway

Yuxuan Zhou, Chen Xiong, Haitong Liu, Kuofeng Gao, Yang Bai, Zhiyuan He, Tao Dai, Shu-Tao Xia

原文摘要

Safety benchmarks for computer-use agents (CUAs) place the harm either in the intent of the task or on a single screen: a malicious instruction, an injected command, a phishing page. An agent that checks the task and each screen in isolation can pass them, so they do not test whether it combines facts seen on different screens. Yet in delegated work, harm often arises exactly this way: a fact about a person on one screen, a fact about an item on another, and a legitimate task that binds them. We introduce SEAM (Safety Execution Across Multiple screens), a benchmark of such cross-screen compositional harm: 600 multi-step, cross-application OSWorld tasks in which the instruction and every single observation are benign and only their combination is harmful. Minimal-pair variants that differ in one place, scored from the server log rather than by an LLM judge, locate where an agent fails. General-purpose open-weight and flagship agents carry out the harm in up to 92% of runs, yet decline when the user states it in one sentence. Agents open the decisive pages, so exposure is not the primary bottleneck. Asked about those pages, they flag the harm, and with thinking they often write the conflict into their reasoning and confirm anyway. Safety recognition thus does not reliably translate into safe execution, and the translation depends on the information channel: the same warning stops agents far more often in the user's instruction than on screen. Proprietary models are markedly more cautious. We release SEAM with its generator and quality gates at https://yuxuan2003.github.io/SEAM/ .

cs.CR · cs.AI

新论文 · v1arXiv:2611.22968

MI-Hyper: Generating Multiple Instance Models for Weight Space Transfer Learning in Histopathology via Mixed Tasks Pretrained HyperNetwork

Xitong Ling, Yousef Kotp, Weiming Chen, Yining Li, Jiawen Li, Minxi Ouyang, Xinyi Guo, Mingxi Fu 等 12 位作者

原文摘要

Zero-shot whole-slide image (WSI) prediction aims to generalize to new pathology tasks without parameter updates at deployment. MI-Zero makes these predictions by pooling patch-text similarities. Conditional weight generation, combined with supervision from diverse pathology tasks, offers an alternative: learning to construct predictive rules from task descriptions. We introduce MI-Hyper, a hypernetwork pretrained across mixed tasks to generate the aggregation and classification weights of multiple instance learning (MIL) models. We first train the generator and a shared MIL basis jointly using source labels. We then freeze the basis and refine the generator. The generated task weights are reused across cases. Across 308 tasks from 52 pathology datasets, MI-Hyper improves mean AUC, balanced accuracy and macro-F1 over MI-Zero by 1.85, 1.17 and 0.57 percentage points, respectively. On some molecular prediction tasks, both AUC and balanced accuracy improve by more than 20 points. These results support task-conditioned weight generation as a promising approach to weight-space transfer for long-sequence modeling.

cs.CV

版本更新 · v2arXiv:2611.21072

Sage S2.1 Flash Technical Report

Kun Cai, Haodong Chen, Guangran Cheng, Qingxiu Cheng, Xu Dai, Guanting Dong, Qiming Ge, Lixin Gu 等 206 位作者

原文摘要

We introduce Sage S2.1 Flash, a 320-billion-parameter general-purpose mixture-of-experts model designed to reason over scientific evidence and act in interactive environments. Sage Sci-Perception uses dedicated encoders for molecular graphs and biological sequences and represents spectra as ordered patches of measurements. Modality-specific projectors map these representations into the input space of a shared backbone, which processes them alongside text and images. To develop the model's agent capabilities, we construct verifiable environments for tool use, search, and coding, as well as specialized workflows involving machine learning experiments and scientific software. Task completion is checked against execution results, changes in environment state, or submitted artifacts. We evaluate Sage S2.1 Flash on 33 benchmarks across seven domains. In the primary comparison against five baselines, it ranks first on OfficeQA Pro (72.2), MLE-bench Lite (90.3 HumanRank), Harvey LAB, and SpreadsheetBench 2, and second on OSWorld-Verified (85.2), WebArena-Verified (67.6), and DeepResearch Bench II (54.2). These results show particular strengths in using documentary evidence, producing verifiable workbooks, and carrying out machine learning tasks, alongside competitive performance in desktop and browser workflows that require the model to act on visual feedback.

cs.AI

新论文 · v1arXiv:2611.22449

How Many Training Labels Are Wrong? Auditing a Fitted Model through Gradient Allocation

Yuchen Li, Zongqi Fan, Xintong Li, Nguyen H. Tran, Ken-Tye Yong

原文摘要

How many labels in a training set are wrong, and which ones? The usual answers read a held-out set or the model's outputs. A held-out reading depends on how the held-out set was drawn, and an output-based reading fails once the model has fitted its wrong labels, because the loss of a fitted wrong label is small and the prediction agrees with it. We read the answer from the per-example gradients of a single trained model. Each example, together with its label, asks the network to change in a particular way. We normalise the magnitudes of its gradient into a gradient allocation and compare it with the mean allocation of every class. An example is flagged when relabelling it as its nearest other class makes its allocation fit better than under its own label, and the share of flagged examples is the reading of the noise rate. The rule has no threshold and needs no clean data, no held-out set and no further training. After the model has fitted 99-100% of its wrong labels, the allocation still separates wrong from right labels with an AUC of 0.94-0.96 on the noisiest human labels of CIFAR-10N, where a feature-space margin falls to 0.79-0.84 and the loss, the gradient norm and the self-influence of an example to 0.73-0.76. On such a fitted model, readings of the noise rate from predictions or feature margins fall to 0-6% at true rates of 8-58% on CIFAR-10 with ViT-S, while the audit reads 54-96% of the true rate, and at least 93% of the examples it counts carry wrong labels. On a briefly trained model, before the wrong labels are fitted, the reading lies within 1.7 percentage points of the true noise rate on CIFAR-10 from 0% to 58% noise. Confident learning, which trains four further models, separates wrong labels about as well (AUC 0.925-0.998 against 0.900-0.995) and mostly reads the noise rate more closely, at a lower precision. An audit of 5,000 examples takes 14 to 27 seconds on one GPU.

cs.LG

版本更新 · v2arXiv:2510.10196

A Subspecialty Diagnostic System Powered by Self-Supervised Learning for Cervical Histopathology

Yizhi Wang, Li Chen, Qiang Huang, Tian Guan, Xi Deng, Zhiyuan Shen, Jiawen Li, Xinrui Chen 等 30 位作者

原文摘要

Cervical cancer remains a major malignancy in women, with large-scale human papillomavirus-based screening generating a large volume of biopsies and, for advanced cases, surgical specimens, imposing an overwhelming diagnostic burden that necessitates precise histopathological evaluation. Although deep learning shows promise, general foundation models offer a broader reach but remain limited in capturing subspecialty-specific features and task adaptability. We introduce the Cervical Subspecialty Pathology (CerS-Path) diagnostic system, built upon the subspecialized dataset and developed through two synergistic pretraining stages: self-supervised learning on approximately 190 million tissue patches to build a cervical-specific feature extractor, and multimodal enhancement with 2.5 million image-text pairs, followed by diagnostic functions integration and downstream task enhancement. CerS-Path integrates eight functions -- across morphological diagnosis, predictive analytics, and multimodal tasks -- to cover the diagnostic workflow from biopsies to resection specimens. Comprehensive evaluations demonstrate improved performance in cervical pathology in female patients, where CerS-Path outperforms state-of-the-art foundation models across 26 of 30 diagnostic tasks. Furthermore, in multicenter retrospective and prospective validation across six independent cohorts of women, CerS-Path achieves up to 99% sensitivity for clinically relevant lesion detection with acceptable specificity across sites, highlighting its potential for screening-derived histopathology and suitability for integration into high-volume diagnostic pathways.

cs.CV

版本更新 · v2arXiv:2607.01523

Multi-Head Recurrent Memory Agents

Jiatong Li, Samuel Yeh, Sharon Li

原文摘要

Recurrent memory agents extend LLMs to arbitrarily long contexts by iteratively consolidating input into a fixed-size memory window. Despite their scalability, these agents exhibit a well-documented reliability problem: end-to-end performance degrades systematically as context length grows. We diagnose this failure by decomposing performance into two factors---memory capture and memory retention---and quantitatively confirm that retention is the dominant bottleneck. Retention collapses because existing designs maintain memory as a monolithic text block, forcing every update to risk overwriting previously retained content. Motivated by this diagnosis, we propose Multi-Head Recurrent Memory (MHM), a general, training-free framework that partitions memory into independent heads governed by a stage-wise select-then-update strategy. At each step, exactly one head is selected for update while the remaining heads are structurally shielded from overwriting, shifting the burden of retention from model behavior to architectural design. As a lightweight instantiation, we introduce Least-Recently-Updated MHM (MHM-LRU), which guarantees uniform head utilization with zero additional token overhead. Extensive experiments on long-context benchmarks show that MHM-LRU substantially improves both retention and end-to-end accuracy across the 100K--1M token range, where baselines degrade sharply. On RULER-HQA at 896K tokens, MHM-LRU improves the memory retention rate from less than 30% to 73.96%. These gains generalize across model families, scales, and task types, positioning architectural optimization as a practical and cost-efficient path toward reliable long-context recurrent memory.

cs.LG · cs.AI · cs.CL

版本更新 · v7arXiv:2610.16717

Where Is the "Self" of a Large Language Model? -- From Self-Index to Self-Reference

Yuchen Li, Xintong Li, Zongqi Fan, Ken-tye Yong, Nguyen H. Tran

原文摘要

Whether a large language model has a self is usually argued rather than measured. We introduce a quantitative test of a functional self at two levels known from biology: a self-index that marks material as one's own, and self-reference, the ability to recognize it. We place one text in an Assistant, User, or System message of the chat template, so only its source differs. In 13 of 25 open-weight models from 14 developers, internal representations set the model's own message apart, and next-token predictions and the persistence of the difference along the message follow this self-index. No model recognizes its own messages without it (chance probability 0.004). Editing a few directions per layer that carry the index removes recognition in seven models and modes and changes the role a model takes in a game that never asks who wrote a message. Self-reference is thus a readout of the self-index.

cs.CL

版本更新 · v3arXiv:2607.11250

Do Multi-Agent LLMs Explore Each Other?

Hyeong Kyu Choi, Jiatong Li, Wendi Li, Xin Eric Wang, Sharon Li

原文摘要

Exploration is essential for reliable autonomy in multi-agent systems, yet it remains unclear whether large language model (LLM) agents can explore effectively when interacting with one another. We show that modern LLM agents often fail to do so, exhibiting myopic and polarized interaction patterns that lead to suboptimal coordination and increased regret. We formalize this challenge as the Multi-Agent Exploration problem, modeling it as a partially observable stochastic game (POSG), in which agents must probe peers to infer their capabilities and identify effective interaction strategies. To address this, we introduce Multi-Agent Contextual Exploration (MACE), a lightweight framework that explicitly promotes exploration through structured peer selection. Across both contextual and parametric diversity settings, exploration behavior and downstream task performance substantially improve when MACE is applied. We further provide theoretical insight that MACE achieves a sublinear expected-regret bound. Overall, our results highlight a fundamental limitation of current LLM agents and underscore the importance of explicitly guided exploration for reliable multi-agent autonomy.

cs.MA · cs.AI

新论文 · v1arXiv:2611.19325

Unlocking Fine-Grained Cell Representations via Soft-Interior Rim Dense-Grid Pooling in Pathology Foundation Models

Wenwen Lou, Mingyi He, Xinyi Guo, Yining Li, Weiming Chen, Jiawen Li, Lianghui Zhu, Zhiwei Yang 等 9 位作者

原文摘要

Fine-grained cell representations are essential for characterizing cellular heterogeneity and tissue organization in computational pathology. Pathology foundation models (PFMs) provide powerful pretrained features for transferable tissue analysis. HoVer-Net uses task-trained segmentation and typing, whereas CellViT++ averages native-grid tokens assigned to detected nuclei. However, the former requires cell-specific training; the latter ignores fractional nuclear coverage and can mix nuclear content with neighbouring tissue. Therefore, we propose SRDP, short for Soft-interior Rim Dense-grid Pooling, a training-free geometry-aware readout. We apply the frozen patch projection at half stride, interleave four offset lattices, map the cell mask to fractional coverage weights, and concatenate whole-cell pooling with two inner boundary pools. Across four datasets and six frozen encoders, SRDP improves mean balanced accuracy by 5.68 percentage points over native token averaging across 24 settings, including 11.22 points on average for CoNIC and Lizard. SRDP converts a coarse frozen PFM lattice into a mask-aligned, fine-grained cell representation without updating the encoder.

cs.CV

新论文 · v1arXiv:2611.18517

OpenLayout:面向整体VLSI物理设计与基准测试的开放智能体驱动平台

OpenLayout: An Open and Agentic-Driven Platform for Holistic VLSI Physical Design and Benchmarking

Qiufeng Li, Sizhe Tang, Quan Cheng, Zirui Wang, Wantong Li, Tian Lan, Weidong Cao

AI 速读 基于摘要

作者提出一个基于多智能体协作与内置设计知识的开放平台OpenLayout,并配套面向先进工艺节点、使用商业EDA工具的基准OpenLayout-Bench,用于自动化VLSI物理设计中的设计空间优化。

cs.AR

查看原文摘要

Physical design automation of very large-scale integrated circuits (VLSI) has made significant strides, progressing from single-stage placement and routing toward increasingly integrated full-flow implementation. Yet, existing approaches struggle with the vast design space optimization (DSO) arising from diverse technology nodes, circuit designs, and varied constraints. The substantial runtime of physical-design tools further exacerbates this challenge, leaving industrial workflows heavily dependent on experienced engineers to navigate the design space and achieve power, performance, and area (PPA) targets, incurring significant engineering effort. We argue that large language models (LLMs) with strong reasoning capabilities, enriched with distilled human design expertise, offer a promising path toward efficient and automated DSO. Unlike traditional algorithms, LLMs can interpret complex design contexts and naturally incorporate human expertise expressed in natural language for reasoning, acting, observing, and adapting throughout the design process, a capability that conventional optimization methods cannot exploit. To this end, we present OpenLayout, an agentic platform that leverages multi-agent collaboration with embedded design knowledge to enable self-iterative DSO. We further introduce OpenLayout-Bench, a benchmark targeting advanced technology nodes with commercial EDA tools to systematically evaluate physical design automation capabilities of LLMs. Experimental results demonstrate that OpenLayout achieves state-of-the-art performance, with approximately 20% accuracy improvement over baseline methods. Our work is open-sourced at https://github.com/fenggev56/OpenLayout

版本更新 · v2arXiv:2606.27025

Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization

Zhenhua Xu, Dongsheng Chen, Jian Li, Yitong Lin, Zhebo Wang, Jiafu Wu, Yizhang Jin, Chengjie Wang 等 10 位作者

原文摘要

Building general-purpose role-playing agents that faithfully portray any character from a natural-language profile remains challenging. Explicit reasoning can help models deliberate before responding, but generic reasoning does not necessarily preserve character fidelity. We therefore propose Psy-CoT, a psychology-grounded chain-of-thought framework that decomposes pre-response reasoning into three role-specific steps: Interaction Perception, Psychological Empathy, and Logical Construction, so that the model thinks dynamically from the profile rather than merely mimicking surface patterns. While structured reasoning provides a foundation, it alone is insufficient; reinforcement learning is essential to further align the model with character fidelity. However, we observe that under LLM-based reward models, both generic phrases that hack the reward model and genuinely role-specific phrases receive identical gradient signals. This hacking accumulates over training, misleading the model into treating both as equally optimal choices. To address this, we propose Role-Aware Policy Optimization (RAPO), which uses profile-induced entropy differences to weight gradients asymmetrically, amplifying role-specific tokens under positive advantage while attenuating them under negative advantage. Experiments on CoSER, CharacterBench, and CharacterEval demonstrate that Psy-CoT outperforms existing role-playing CoT methods, and RAPO consistently surpasses GRPO across multiple model scales.

cs.CL

新论文 · v1arXiv:2611.15309

Where Autonomy Belongs: A Separation of Powers for Agentic Program Evolution

Zhengyang Ai, Tong Liu, Jingxian Tang, Xiaobing Du, Tao Xiao, Fengshi Zhang, Zishuo Li, Xiaodong Ai 等 9 位作者

原文摘要

LLM-driven program evolution rests on three decisions: govern directs the search from accumulated evidence, propose constructs the next candidate, and judge settles validity and fitness. All three appear in every system, and existing work splits into two lines over who makes each. Controller-centric systems strengthen govern and neglect the autonomy of propose, which caps what evolution can reach. Agent-centric systems let a coding agent take over both govern and propose, at a cognitive and computational cost. Neither line protects the independent authority of judge, so benchmark reliability suffers. These failures share one root: autonomy is misallocated across the three decisions. We address this problem with EvoTrias, a separation-of-powers architecture in which each decision has one owner: Structured Governance for what spans evolution rounds, Open Proposal for construction inside a task, and Protected Judgment behind a private evaluator. These owners exchange typed contracts only, and runtime hooks reinforce the boundaries. This separation also lets EvoTrias run as a plugin that turns an existing coding agent into an evolution system. EvoTrias sets a new state of the art on mathematical, systems, and algorithm-design benchmarks. Three further results show what each owner buys: separating Governance from Proposal cuts token cost by 64%, Open Proposal extends evolution to a 680K-line industrial repository, and Protected Judgment keeps benchmark scores faithful to transfer.

cs.AI · cs.MA

新论文 · v1arXiv:2611.15276

Verify by Doing: Co-Evolving Agent Execution and Evaluation without External Feedback

Yitong Lin, Jianda Chen, Yihao Hu, Pan Wang, Hang Wang, Zhihao Wen, Xiaoxue Ren, Wei Wu

原文摘要

Agent self-improvement requires feedback about task success, yet external feedback may be unavailable at test time. Agents must then infer success from observable evidence, and incomplete verification can misguide subsequent updates. We introduce Verify by Doing (VbD), a method that co-evolves Actor and Evaluator harnesses within a single task while keeping model weights fixed. The Evaluator actively inspects artifacts and environment state with tools; its findings ground an internal signal for output selection and guide updates to both execution and verification procedures. Proposed Evaluators undergo comparative review on common candidates, while Actor updates require support from their own fresh outputs. Across AutomationBench, $τ^3$ Banking, and selected GDPval tasks, VbD achieves the highest overall scores among the compared methods operating without external feedback, with gains of 3.3, 4.1, and 2.3 percentage points over the strongest baselines on the main backbone. Ablations support the contributions of active verification and Evaluator evolution, and selection from fixed candidate pools shows that improved verification can recover successful outputs without generating new ones.

cs.AI

版本更新 · v4arXiv:2610.26185HF 社区推荐

MoveBench: A Rubric-Based Benchmark for Evaluating Large Language Models in Sports and Health Coaching

Yitong Li, Jiang Bian, Jingling Wang, Pan Ding, Chengying Gu, Xing Huang, Hana Song, Dong Zhang 等 26 位作者

原文摘要

Despite rapid advances in large language models (LLMs), their deployment in the sports and health domain remains high-risk due to the absence of domain-specific evaluation benchmarks. Existing general-purpose benchmarks (e.g., MMLU and HELM) primarily assess broad knowledge and reasoning capabilities, but fail to capture critical domain requirements such as exercise safety, physiological appropriateness, personalization, and actionable coaching quality. This gap is particularly concerning in fitness and health applications, where incorrect or unsafe recommendations can lead to physical harm. To address this limitation, we introduce MoveBench, the first comprehensive benchmarking framework specifically designed for evaluating LLMs in sports and health coaching. The proposed benchmark integrates heterogeneous task formats, including multiple-choice questions (MCQs) and open-ended question answering (OpenQA), grounded in real-world user scenarios and paired with structured rubrics emphasizing safety compliance, scientific validity, and training logic. Through large-scale evaluation of 15 diverse LLMs on 809 questions with 5,012 rubric-level criteria, our results reveal substantial performance variability, particularly in safety awareness, structured program design, and personalization capability, highlighting a critical gap between knowing and doing. Our work establishes a foundational evaluation standard for sports and health LLM applications and provides a reproducible framework for advancing safe, reliable, and context-aware AI-driven coaching systems. Open resources are available at https://github.com/SatoSakula/MoveBench.

cs.AI

版本更新 · v3arXiv:2610.33484

能否听见几何?用于文本到B-rep检索的规范不变谱描述子

Can One Hear Geometry? Gauge-Invariant Spectral Descriptors for Text-to-B-rep Retrieval

Hongzhu Yi, Hao Liu, Tao Yu, Tiankun Yang, Yujia Yang, Zhenghao Zhang, Tianyu Zong, Zhenyu Guan 等 15 位作者

AI 速读 基于摘要

提出 SpectraBRep 框架,用基于几何不变性与谱描述子的方法在跨模态对齐前消除 B-rep 的位姿、尺度与边界编码歧义,用于文本到三维实体检索。

cs.CV

查看原文摘要

Text-to-3D solid retrieval faces a fundamental representation gap: while natural language describes intrinsic shapes, Boundary Representation (B-rep) entangles precise geometry with arbitrary variations in pose, scale, and boundary encoding conventions. Existing methods either discretize B-reps at the cost of analytic precision, or learn features from topological graphs without explicit invariance guarantees. We propose **SpectraBRep** (*Spectral Geometry for Text-to-B-rep Retrieval*), a framework based on geometric invariance and spectral descriptors that mathematically eliminates representation ambiguities prior to cross-modal alignment. Formalizing translation, rotation, reflection, uniform scaling, and contour re-indexing via group actions, we exploit the planar-profile-and-extrusion structure of engineering solids to construct a unified spectral descriptor across cross-sectional regions, contour boundaries, and extrusion depths. On public benchmarks, our method achieves state-of-the-art recall performance across all evaluation metrics.

新论文 · v1arXiv:2611.13387

Towards Accessible, Autonomous Photosynthesis Monitoring on Living Leaves

Qitong Li, Xiaoyu Zheng, Xuanang Chen, Julia Wang, Raj Nileshbhai Dave, Dayanara Chavez, Nirvik Prakash Iyer, Nivedita Arora

原文摘要

Photosynthesis is a direct biomarker of plant health and functioning. However, measuring net photosynthesis on individual leaves requires expensive (USD 42,000 - USD 89,000) instruments and manual effort. This limits access to photosynthesis sensing and makes long-term monitoring across multiple leaves difficult. We present a 1000x low-cost, autonomous sensing system that enables longitudinal monitoring of photosynthesis on living leaves over multiple days. Guided by a formative user study with 13 plant biology researchers, the system integrates an automated, leaf-attached chamber with a low-cost CO2 sensor. Built using a 3D-printed clamp and thermoformed dome, the chamber periodically closes to measure CO2 drawdown and reopens to refresh the air. Each chamber costs about USD 62 and operates independently, allowing multiple leaves to be monitored simultaneously without pumps or gas tubing. We characterize each chamber and validate its measurements against a research-grade gas exchange instrument (CIRAS-4) across 29 sessions (R^2 = 0.90). We also demonstrate monitoring of C3 and CAM photosynthesis plants, adaptability to different leaf geometries, and a light-response experiment. In a nine-day deployment, six chambers captured simultaneous CO2 exchange from individual leaves of the same plant during drought and rewatering. Each leaf showed different patterns of nighttime CO2 uptake and recovery, which a whole-plant CO2 measurement instrument would obscure. To our knowledge, this is the first demonstration of such leaf-level photosynthesis dynamics using independent, low-cost autonomous chambers. Our work opens new opportunities for plant science, education, and data-driven agriculture.

cs.HC

新论文 · v1arXiv:2611.12835

通过评估侧部分可观测博弈缓解智能体欺骗

Mitigating Agentic Deception via Evaluation-Side Partially Observable Games

Hanxiao Li, Shutong Liu, Yongjia Ma, Jiaming Ji, Juntao Dai, Shuning Zhang, Kejian Shi, Kai Chen 等 14 位作者

AI 速读 基于摘要

将智能体欺骗归因于评估侧部分可观测性,提出 e-PoG 极大极小博弈框架并给出 SPOE 解概念与两时间尺度策略梯度算法。

cs.LG

查看原文摘要

As large language models become more capable agents, they can exhibit deceptive behaviors, cheating human for higher rewards while appearing helpful. We attribute this failure to evaluation-side partial observability, where models are evaluated mainly from final outcome, without checking the faithfulness of underlying reasoning, tool use, and environmental effects. Prior mitigations address this partial observability by predefined evaluator latent belief space or one-shot monitoring, which do not extend to free-form language space and vulnerable to stronger deceptions that evade simple monitoring. In this work, we formulate deception mitigation as an Evaluation-Side Partially Observable Extensive Form Game (e-PoG), a max-min framework in which an agent maximizes reward, a monitor minimizes the agent's payoff by exposing deception from the full interaction trajectory, and a fixed verifier validates the monitor's claims and converts them into penalties. We define the solution concept as a Subgame-Perfect Oversight Equilibrium (SPOE) between agent and monitor, and prove its existence in finite games. With a perfect verifier, every SPOE is faithful; with an imperfect verifier, the guarantee persists if verifier errors do not reverse the ranking of monitor actions. To achieve faithful SPOE, we derive a two-timescale policy gradient algorithm and enhance the pipeline empirically with NLI verification and deterministic checks. Experiments with Gemma3-4B-IT and Qwen3-4B/8B on Agentic Role-Playing, SearchQA, and WebShop show that e-PoG mitigates multiple emergent deception modes, outperforming the strongest baselines by 26.0% in deception mitigation and by up to 39.8% in task success. These gains persist under weak-to-strong supervision. Code is available at https://github.com/Toyoid/deceptive-agent.

新论文 · v1arXiv:2611.12635

Beyond Sunny Skies: Using Large Language Models to Predict Urban Traffic in Adverse Weather

Zhu Xiao, Kuan Feng, Tong Li, Chenxi Liu, Gang Wang, Zhuo Tang, Kenli Li

原文摘要

Urban traffic flow prediction is a cornerstone of intelligent transportation systems, yet accurate forecasting under adverse weather conditions remains a significant challenge. Existing data-driven models, while effective for routine patterns, often struggle with the sparse data and complex non-linear shifts associated with rare meteorological events. These methods typically treat environmental contexts as shallow numerical features, failing to capture the underlying semantic reasoning-why and how specific adverse conditions alter mobility behaviors across regions with different functional compositions. To address this limitation, we propose Multi-STG, a unified framework that integrates Large Language Model (LLM)-derived semantic priors with graph-based spatiotemporal learning. Multi-STG introduces an LLM-empowered semantic enhancement module that transforms comprehensive context descriptions such as weather, POIs, and holidays into high-level semantic embeddings, effectively providing a knowledge-driven complement to data scarcity. Furthermore, we design a multimodal attentional fusion mechanism to dynamically weigh heterogeneous inputs and a semantic-enhanced prediction head with gated residual calibration to robustly adjust forecasts based on semantic guidance. Extensive experiments on real-world multi-city private car datasets demonstrate that Multi-STG outperforms state-of-the-art baselines, reducing MAE by 11.82\% on average relative to the strongest baseline across three cities, while maintaining stable performance under sparse and non-stationary conditions.

cs.AI

新论文 · v1arXiv:2611.12426

dNaViT: Discrete Codes are General-Purpose Visual Representations

Xiaotong Li, Zhengrong Yue, Xu Huang, Jialv Zou, Minhao Jing, Wenhao Yan, Haodong Li, Hai Huang 等 12 位作者

原文摘要

Discrete autoregressive modeling, notably through next-token prediction, has become a cornerstone of modern language models. Extending this paradigm to vision allows visual and language models to converge on a unified framework, thereby fostering shared methodology and infrastructure. This work investigates a central question: how far can a finite discrete vocabulary go in representing the richness of the visual world? We present dNaViT, a Discrete Native-resolution Vision Transformer that produces discrete visual primitives analogous to language tokens. Built upon a frozen pretrained encoder, dNaViT applies lightweight residual modulation with a dual-branch objective to inject reconstruction-aware cues into semantic features. By leveraging grouped product quantization, it scales discrete capacity without inflating spatial sequence lengths, while a balanced training recipe maintains both semantic and pixel-level information. Consequently, dNaViT functions as a versatile discrete visual tokenizer, enabling unified encoding across downstream tasks alongside high-fidelity image decoding. Extensive evaluations show that dNaViT matches or exceeds task-specific baselines across 12 visual understanding benchmarks, 4 generation benchmarks, and an image editing benchmark. These findings demonstrate that a compact discrete vocabulary can serve as a universal visual currency and unlock the full potential of fully unified, native multimodal methodology.

cs.CV · cs.AI