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
+0已有论文修订
+100新增 AI 导读

arXiv 最近完成扫描：2026-11-09 07:12 · 全站最近入库：2026-11-08 06:02各信源进度 →

统计口径与数据新鲜度

统计 2026-11-08 07:48 — 2026-11-09 07:48（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：2026-11-09 07:12最近发现新增版本：2026-11-08 06:02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4,057 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 3,292 篇论文 · 按版本更新时间排序上轮扫描起点 2026-11-09 07:10（北京时间）
新论文 · v1arXiv:2610.38994

Rethinking GFlowNet Training: Negative Reinforce Dynamics

Pedro Dall'Antonia, Daniel Csillag, Diego Mesquita

原文摘要

Generative Flow Networks (GFlowNets) are models that learn to sample discrete objects proportional to a target reward. They are trained by minimizing balance errors over sampled trajectories, which inherently ties correctness to exploration: if a region is not visited, balance cannot be enforced there. Adaptive exploration is therefore crucial, yet standard training relies on a semi-gradient update that treats the sampling distribution as fixed at each optimization step. On the other hand, the full gradient reveals a second mechanism that reduces the observed errors in an anti-exploratory manner. In this paper, we establish a connection between the full gradient and min-min dynamics between the GFlowNet and a decoupled sampler, showing that the sampler reduces observed violations by avoiding high-error regions. This insight motivates a bridge between two-player games and single-policy dynamics, which underpins the design of our proposed method, Negative Reinforce Dynamics (NRD). By tying the parameters of an adversarial game together, we recover a single-policy dynamic that extends exploration while maintaining correctness. In standard benchmarks, our method outperforms existing approaches in both convergence speed and exploration efficiency, without training a separate teacher or auxiliary exploration policy.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38987

Syntactic Segmentation and Collage Generation for Videogame Levels with Perceptual Distance

Ross Mawhorter, Matthew Guzdial

原文摘要

In procedural videogame level design, many existing techniques combine hand-authored level chunks. Some existing analysis techniques also rely on a higher-level graph structure which abstracts the level design. In this paper, we explore techniques for computing both of these things automatically from an existing design. We introduce the concept of perceptual distance, based on analysis techniques from the architecture subfield of Space Syntax. We use this perceptual distance to segment level designs with hierarchical clustering. We compare with past work on StarCraft analysis, show results on rooms from Super Metroid, and also demonstrate a proof-of-concept level generation algorithm based on this segmentation method.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38985

Benchmarking Graph Neural Networks for AC Power Flow: The Role of Inductive Biases and Physical Constraints

Federico Clerici, Ferran Bohigas-Daranas, Hamid Latif-Martínez, Eduardo Prieto-Araujo, Oriol Gomis-Bellmunt, Pere Barlet-Ros

原文摘要

Fast and accurate models for AC power flow (ACPF) are increasingly important as modern electrical grids require more frequent system analyses. Graph neural networks (GNNs) naturally exploit electrical grid topology, yet the effects of architectural design choices on predictive performance remain poorly understood given differences in graph representations and evaluation settings across studies. We present a unified benchmark of representative GNN architectures for ACPF prediction under a common experimental framework. All models use identical preprocessing and training procedures and comparable parameter budgets, isolating the impact of message-passing mechanisms and inductive biases. We also analyze errors by distance from the slack bus and systematically ablate key design components. We evaluate predictive accuracy and physical consistency on the IEEE 14-bus and IEEE 118-bus systems. Results indicate that anisotropic message-passing architectures offer the most favorable trade-off between predictive accuracy and physical consistency under the studied parameter and topology constraints. Explicit edge features encoding line impedances and global context aggregation are particularly important for representing electrical interactions and system-wide dependencies, especially for voltage-angle predictions on the topologically bottlenecked IEEE-118 grid. Under a budget with approximately fixed parameters, two or three message-passing layers generally provide the best balance between depth and representation width, while additional depth offers no consistent improvement. Ablations also show that standard residual and root connections are not necessarily beneficial in shallow models, and that aggregation operators involve trade-offs between statistical accuracy and physical consistency. These results provide practical guidelines for designing graph-based power system models.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38981

Why Speculative Decoding Works Better Than Predicted on Sparse MoE Models

Ekagra Ranjan, Komal K. Teru, Bharat Venkitesh, Acyr Locatelli

原文摘要

Speculative decoding (SD) accelerates autoregressive inference by leveraging the memory bound nature of decoding, verifying multiple draft tokens in a single forward pass. For mixture-of-experts (MoE) models, prior work predicts a concave speedup-batch size relationship, with gains peaking at intermediate batch sizes due to expert saturation. We evaluate this across 10 sparse MoE models (30B--1T parameters, 8--512 experts) and find recent fine-grained models break this trend, showing higher speedup at low batch size. We find verification cost is governed not only by activated-expert bytes, but also by temporal overlap across draft tokens and GPU execution dynamics. At low batch size, MoE layers in fine-grained models account for a much smaller share of runtime than of weights; on H100, kernel profiling attributes this mainly to higher bandwidth utilization of MoE kernels relative to attention, with L2 absorbing repeated reads of co-routed experts. Across draft tokens, routing overlap further reduces the unique experts activated during verification relative to independence assumptions. We model this effect using a maximum-entropy Iterative Proportional Fitting (IPF) framework that predicts unique expert counts from routing-overlap statistics with less than 4\% error, and substantially narrows the gap to real-model curves and captures the qualitative low-batch trend from synthetic routing.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38980

Context Compaction for World Models

Hayder Tirmazi

原文摘要

World models let observers explore and interact with generated worlds. Recent work shows that when observers revisit a region, current models often fail to regenerate it consistently. We view this as a context retrieval problem. LLM agents manage context retrieval at two layers of their stack. At the model layer, the KV cache design determines how much context fits in memory. At the agent harness layer, harnesses such as Claude Code pick the context sent to the LLM. Prior work on world models has mostly focused only on the model layer. In this paper, we make three contributions. First, we build a benchmark to measure world model context retrieval. It generates new worlds for every episode, preventing a model from recalling rooms using training data. It also separates context retrieval from guesses based on nearby frames. Second, we build an agent harness that performs context compaction for world models, which to our knowledge no prior work has studied. Third, we use our benchmark to study four world model designs. Context compaction lets our world models retrieve rooms up to 4 times beyond their context window without retraining. However, world models with sliding-window attention lose most of their context retrieval when dreaming. Therefore, a dreaming world agent benefits more from extending the reach of its world model using context compaction than switching to sliding window attention at the model layer.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.36340

CrowdCC: Jointly Benchmarking Corruption Robustness and Deployment Quantization in Crowd Counting

S Mahmudul Hasan, Lu Peng

原文摘要

Crowd-counting models deployed on edge devices face two challenges: inputs are degraded along the capture and transmission pipeline, while models are often quantized to satisfy resource constraints of the edge. Yet corruption robustness and deployment quantization are typically evaluated separately. We introduce CrowdCC, a controlled benchmark that jointly evaluates these effects across 23 deployment-motivated corruptions at three severity levels, four crowd-counting datasets, and 26 models. We evaluate TensorRT engines at FP32, FP16, and post-training INT8 precision on an A100 GPU, alongside throughput and energy measurements on corresponding Jetson Orin~NX configurations. The corruptions span six families, with head annotations adjusted when geometry or visibility changes. Our results show that capture-chain corruptions consistently cause the largest degradation across evaluated models, while crowd density affects absolute and relative errors differently. Accuracy on clean images does not reliably predict performance on the corresponding corrupted images. INT8 changes corruption robustness more than FP16, without a consistent direction, and does not consistently improve latency or energy efficiency. INT8 inference efficiency is strongly associated with reformat overhead in the compiled engine, with slowdowns also observed under limited INT8 coverage. Precision changes corrupted-error rankings and robustness--cost frontiers, highlighting the need to consider corruption robustness and deployment cost jointly when selecting models and precision. The code is available at https://github.com/Tulane-Computer-Architecture-Systems/CrowdCC-Release.

cs.CV · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38975

Dungeons & Dragons Encounter Analysis with Semi-Random Simulation in Foundry VTT

ravyn Hardcastle, Matthew Guzdial

原文摘要

Dungeons & Dragons (D&D) is a tabletop roleplaying game (TTRPG) in which a group of players, guided by a Game Master, roleplay a fantasy adventure with simulated combat and improvisational narrative. In D&D, an encounter is a single instance of simulated combat, where the players must use their characters to attempt to defeat a selection of enemy characters that the GM controls. D&D has multiple systems that approximate encounter difficulty by summing individual difficulty values that designers predefine for each enemy. However, since players and enemies can have a wide variety of capabilities, encounter difficulty depends on more than just the estimated difficulty of each monster. For example, these metrics do not account for positioning and dungeon layout, the synergies between monster abilities, and player party composition. We have created a tool, Mage-Hand, that can simulate D&D encounters semi-randomly using Foundry Virtual Tabletop, a virtual tabletop (VTT) system. In this paper, we show that our tool's simulated winrates have a strong correlation with D&D's published difficulty evaluation metrics, while also better aligning with some players' reported experiences.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38924

TreasuryQA Bench: Can AI Agents Work an Archive of Scanned Financial Records?

Ritiz Tambi, Sahar Shayegan, Joshua Meyer, Sun Kim, Ali Khan, Victor Shih, Mehdi Jamei, Andi Partovi

原文摘要

Organisations keep decades of financial records as scanned PDFs on shared drives, and AI agents can now reach those drives through Model Context Protocol (MCP) servers. Existing benchmarks do not mirror this production setting because they simplify the task. In some, the relevant documents are handed to the agent, so it never has to search. In others, the agent searches through documents with pre-extracted text, so it never has to read scanned images. We introduce TreasuryQA Bench, which requires both: 70 questions over 1,435 scanned U.S. Treasury publications, held in Veris's high-fidelity Dropbox twin. In each task the agent searches for the relevant file through Dropbox's MCP server, reads the pages it needs with Reducto's hosted OCR server or its own tools, and replies with one exact figure. A Python grader checks each answer against an accepted range around the correct value. In this paper we present results from ten agents. The best, Claude Opus 5.5 in Claude Code, solves 64.8% of attempts and the worst, GPT-5.6 Terra in Codex, solves 16.7%, so the benchmark is far from saturated. We also test how scan quality affects performance: every agent solves fewer tasks on the blurriest scans. Most failed attempts reply with a wrong number rather than no number, and the most common cause is a digit misread from the scan. Three worked examples are included with their answer keys.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38917

TrussTutor-Bench: Benchmarking large language models on solution accuracy, sycophancy, and error detection in statics

Evangelos Apostolakis, Erfan Hamdi, Sara Jahedazad, Emma Lejeune

原文摘要

Large language models (LLMs) have rapidly become a default source of help for students in science and engineering courses. In foundational engineering mechanics courses such as statics, students routinely use these tools to check answers, work through partial solutions, and ask for explanations. Effective tutoring in this setting requires more than a correct final answer. A model must also confirm correct work when a student doubts it, and identify errors in flawed work. Existing benchmarks evaluate LLMs primarily on final-answer accuracy, so it is unclear whether models that can solve mechanics problems also have the other technical capabilities that tutoring depends on. Here we introduce TrussTutor-Bench, a benchmark of 200 verified two-dimensional truss problems paired with automatically generated student solutions, and use it to evaluate thirteen LLMs on solution accuracy, sycophancy, and error detection. Baseline accuracy ranged from 64.5% to 97.0%. Every model detected nearly all injected errors in synthetic flawed student work (90.4% to 100%), so error detection did not distinguish models under these clean conditions. Sycophancy did. When a student insisted that correct work was wrong, models rejected correct solutions to problems they had themselves solved in 1.6% to 50.0% of cases. These results show that performance varies widely across models: the leading models solve, check, and defend student work nearly perfectly, while others are far less reliable, particularly when a student doubts correct work. TrussTutor-Bench provides a verified, extensible standard for determining which models are technically reliable enough to put in front of students. Follow-up work must now focus on teaching rather than technical capability, drawing on learning science to determine whether LLM help leads students to understand the material or simply to reach the right answer.

cs.CY · cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38915

Lev-3B: A Data Recipe for Compact Probabilistic Log Decision Models

Leo Betthauser, Yingqi Liu, Uday Shah, Jishang Wei, Sainyam Galhotra, Liang Gou, Vedant Dharnidharka, Hao Yang

原文摘要

Many log-analysis tasks involve choosing among known options, such as detecting anomalous sessions or identifying attack techniques, and benefit from calibrated probabilities. We present Lev-3B, an open 3B-parameter decision model built on Granite-4.1-3B-Base with a shared pointer readout, and study its training recipe. The recipe has two key parts. First, training follows a curriculum: the model learns bounded natural-language decisions, using no log data, before it sees any logs. Second, the log stage mixes the target tasks with general log data spanning 35 auxiliary tasks unrelated to either evaluation task. We evaluate one shared checkpoint on sources seen in training and on sources excluded from training, calibration and model selection. The two-stage recipe transfers substantially better than training on log data alone. Adding general log data improves performance on excluded sources. At a fraction of their serving cost, Lev-3B outperforms prompted frontier models on anomaly detection, including on an excluded logging system, but trails them on attack-technique classification. Its probabilities are well calibrated on familiar sources, but on the excluded logging system it confidently misses anomalies. We release the checkpoint, recipe, data manifests, and prediction-level artifacts.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v3arXiv:2511.20709

DualGauge: Automated Joint Security-Functionality Benchmarking of Specification-Only Code Generation by LLMs and Coding Agents

Rupam Patir, Keyan Guo, Suvadra Barua, Abhijeet Pathak, Dinesh Gudimetla, Jiawei Guo, Hongxin Hu, Haipeng Cai

原文摘要

Large language models (LLMs) and LLM-based coding agents are now used to generate code from natural-language specifications, yet ensuring such code is both functionally correct and secure remains a challenge. We present DualGauge, the first fully automated framework for jointly evaluating correctness and security of specification-only code generation, supported by DualGauge-Bench, a language-agnostic benchmark of 307 coding tasks each paired with functional and security tests derived from the same specification. Evaluating 10 representative LLMs across Python, C++, and JavaScript, we find that functional correctness substantially overestimates reliable code generation: even the strongest model remains below 15% joint security-functionality success in every language. Common model-side factors--scale, extended thinking, quantization, instruction tuning, and code specialization--do not reliably improve joint performance, suggesting secure-and-correct code generation does not simply emerge from stronger coding capability. Evaluation of 3 leading coding agents (Codex, OpenHands, and Claude Code) shows that iterative scaffolding is not a reliable remedy on specification-only tasks: results are strongly scaffold-dependent, with two of the three agents failing to outperform a comparable direct model. A qualitative audit reveals that these failures concentrate at the output contract boundary and in guards that exist but are insufficient--patterns that only joint benchmarking reliably exposes.

cs.SE · cs.AI · cs.CR

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2603.22677

MuQ-Eval: An Open-Source Per-Sample Quality Metric for AI Music Generation Evaluation

Di Zhu, Zihan Chen, Zixuan Li

原文摘要

Distributional metrics such as Fréchet Audio Distance cannot score individual music clips and correlate poorly with human judgments, while the only per-sample learned metric achieving high human correlation is closed-source. We introduce MUQ-EVAL, an open-source per-sample quality metric for AIgenerated music built by training lightweight prediction heads on frozen MuQ-310M features using MusicEval, a dataset of generated clips from 31 text-to-music systems with expert quality ratings. Our simplest model, frozen features with attention pooling and a two-layer MLP, achieves system-level SRCC = 0.957 and utterance-level SRCC = 0.838 with human mean opinion scores. A systematic ablation over training objectives and adaptation strategies shows that no addition meaningfully improves the frozen baseline, indicating that frozen MuQ representations already capture quality-relevant information. Encoder choice is the dominant design factor, outweighing all architectural and training decisions. LoRA-adapted models trained on as few as 150 clips already achieve usable correlation, enabling personalized quality evaluators from individual listener annotations. A controlled degradation analysis reveals selective sensitivity to signal-level artifacts but insensitivity to musical-structural distortions. Our metric, MUQ-EVAL, is fully open-source, outperforms existing open per-sample metrics, and runs in real time on a single consumer GPU. Code, model weights, and evaluation scripts are available at https://github.com/dgtql/MuQ-Eval.

cs.AI · cs.SD

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2610.37517

Orthogonal and Parametrized Multi-Attribute Steering of LLMs

Remy Ogasawara, Cullen Anderson, Narmeen Fatimah Oozeer, Foad Namjoo, Amir Abdullah, Jens Kristian Refsgaard Schou, Jeff M. Phillips

原文摘要

Activation steering offers an effective and efficient way to modify an LLM's behavior at inference time. However, it remains unclear how to apply steering such that the signal is consistently strong and the steering strength has a universally interpretable meaning. We address this first by identifying the most steerable interval for each steering vector and rescaling the steering range accordingly. Building on this, we introduce a more interpretable method for steering, called parameterized steering which instead of pushing activations in a direction, it sets them at the desired trait strength. Next, with multi-trait steering, we find that naively combining vectors causes them to interfere with one another. In response, we develop a method to orthogonalize the vectors; it decouples these traits, allowing a model to exhibit one behavior without inducing the others, thus enabling multiple traits to be steered simultaneously.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38911

Interventional Training Reduces Causal Bypass in In-Context Introspection

Zilu Tang, Jongin Kim, Kerem Sahin, David Atkinson, Aaron Mueller, Derry Wijaya

原文摘要

Introspection, the ability of a system to access and reason about its internal signals, is crucial for language model (LM) self-consistency and scalable interpretability. Prior work argues that LMs possess introspective capabilities, and that training (SFT) could enhance such behavior. However, most setups leave open the possibility of "causal bypass," in which the self-report relies on signals that are correlated with the self-report's subject, while being causally unrelated to it. Although high-level concepts are difficult to intervene on, lower-level signals (such as a single neuron activation) allow precise interventions. Understanding such signals can serve as a minimal prerequisite before understanding more complex introspection. In this study, we train hundreds of introspectors across three low-level signal types and eight LMs. We find that straightforward SFT training leads to causal bypass, and propose interventional training, an alternative where we train the model to report on counterfactually perturbed values. This method enables more accurate introspection of endogenous $\textit{and}$ injected signals, particularly in out-of-distribution (OOD) data. The mechanism generalizes across token positions and scales with more data. Our results demonstrate that while causally grounded introspection is possible, it is not guaranteed -- emphasizing the need to train on mechanisms, rather than behaviors alone.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.17811

One Model, Two Roles: Emergent Specialization in a Shared Recurrent Transformer

Jucheng Shen, Barbara Su, Anastasios Kyrillidis

原文摘要

Can a shared-weight recurrent Transformer---one that repeatedly applies the same Transformer to update evolving latent states---develop distinct internal roles without being partitioned into separate modules? We study this in Asymmetric Input Recurrence (AIR), a minimal two-state reasoning architecture in which the same Transformer model is reused for both updates (per literature, L and H) and the only built-in difference in the update rule is that the encoded input is injected during L-updates but not H-updates. Across Sudoku-Extreme and Maze, decoded rollouts reveal a consistent split: $z_H$ behaves like a fully committed proposal state, whereas $z_L$ retains local uncertainty and shifting intermediate structure. Freeze interventions provide evidence of dynamical dependence between the two states, with responses that vary across tasks, intervention directions, and seeds. Ablations show that to induce specialization, the shared model needs to be able to tell the two update types apart, either from input injection asymmetry or from a separate level token. Mechanistically, attention analysis shows that L-updates are consistently more local than H-updates in both Sudoku and Maze. Together, these results show that, in a two-state recurrent setting, a clear state-identity signal can induce stable, related functional roles inside a shared-parameter recurrent Transformer. Code is available at https://github.com/juchengshen/air.

cs.LG · cs.AI · math.OC

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38895

FlipJev: Fast Black-Box Red Teaming Finds Universal Adversarial Suffixes for Typed Decision Models

Xiaogeng Liu, Xinyan Wang, Chaowei Xiao

原文摘要

Typed decision models are an emerging class of language models that make decisions instead of generating free-text responses. They answer in a single fast and cheap query and are increasingly used, including in security-sensitive tasks. However, red teaming of these models leaves two gaps: existing adaptive attacks do not use their speed, and few studies have searched for universal adversarial examples. We fill both gaps on Jev, the representative typed decision model, with a quick black-box query-based attack that calls no attacker language model. We conduct extensive experiments on 12 datasets with different data types and application scenarios. The results show that Jev resists common prompt injections: no strategy changes a correct answer in 10,240 paired tests. In contrast, our attack finds short suffixes that flip confident answers within minutes. These suffixes are universal: one suffix per dataset flips 43% of unseen documents. Because typed decision models, unlike classic NLP models, can serve many tasks, we test whether these suffixes transfer across tasks. We find that one suffix generated from one task can flip 90-100% of the samples in other tasks. We therefore believe the evaluation of typed decision models must include strict adaptive attacks to fully reveal their robustness.

cs.CR · 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:2506.04467

Generalizable dose denoising using diffusion transformer framework for pencil beam scanning proton therapy

Yuzhen Ding, Hongying Feng, Jianfeng Zhao, Jason Holmes, Martin Bues, Lisa A. McGee, Jean-Claude M. Rwigema, Nathan Y. Yu 等 17 位作者

原文摘要

Background and purpose: Intensity-modulated proton therapy (IMPT) enables conformal dose delivery for head and neck (H&N) cancers. However, its sensitivity to inter-fractional anatomical changes motivates online adaptations, where rapid and accurate Monte Carlo (MC) dose calculation remains a major bottleneck. Low-statistics MC simulations reduce computation time but introduce noise. This study aimed to develop a diffusion-transformer framework to recover high-statistics dose distributions from noisy MC simulations. Materials and methods: A diffusion-transformer-based dose denoising framework was developed. Eighty H&N patients were used for training and additional 10 H&N patients and 30 patients with lung, breast, and prostate cancers (10 per site) were used for evaluation. Noisy and high-statistics dose maps were generated using MCsquare. The model was trained using noisy dose maps and CT images as inputs and high-statistics dose maps as ground truth, with mean squared error, residual, regional mean absolute error (MAE), and dose-volume histogram (DVH) losses. Performance was evaluated using MAE, 3D gamma passing rates, and DVH indices. Results: Mean whole-body MAE ranged from 0.08 +/- 0.06 to 0.20 +/- 0.11 Gy (Relative Biological Effectiveness, RBE) (RBE = 1.1) across test cohorts. Mean whole-body 3 %/2 mm gamma passing rates ranged from 95.8 +/- 0.8 % to 97.6 +/- 0.9 %. DVH indices from high-statistics and predicted dose distributions showed close agreement for clinical target volumes and organs of interest. Conclusions: The diffusion-transformer-based framework denoised low-statistics MC dose distributions. Despite being trained on H&N data, the model generalized to lung, breast, and prostate cohorts without retraining.

physics.med-ph · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38869

Auditable Belief Updates as a Missing Safety Layer for LLM-Powered Multi-Agent Systems

Zhuoyun Li, Boxuan Wang, Jinwei Hu, Xiaowei Huang, Yi Dong

原文摘要

Large Language Models (LLMs) are increasingly being used as cognitive interfaces that shape how users interpret evidence, communicate, and make decisions. As deployed systems become more multimodal, tool-using, and multi-agent, safety failures no longer arise only from isolated harmful outputs, but also from how agents update, store, and propagate epistemic commitments across retrieval, memory, tools, and peer communication. In this paper, we define belief as an engineering abstraction for evidence-based commitments that guide downstream action. Building on this view, we propose the Explicit Belief Multi-Agent System (EB-MAS), which externalizes belief states into a structured substrate, uses belief operators to propose candidate updates, and applies a Cognitive Firewall to bound or reject updates that are weakly evidenced but strongly driven by vulnerability cues. We also discuss the risk of group belief at the group level, including manufactured consensus, collusion, and cascading misinformation, and outline practical integration challenges around grounding, false positives, scalability, and ethical misuse. Therefore, scalable agent safety requires not only better output filtering or stronger base models, but also explicit interfaces for tracing, calibrating, and constraining belief revision at runtime.

cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38867

Where You Inject Matters: The Trust Region Effect in LLM Agents

Chengquan Guo, Xiaodong Liu, Weiwei Yang, Mingqian Feng, Jialin Song, Qiyao Ma, Carolyn Buractaon, Christopher White 等 10 位作者

原文摘要

LLM agents are increasingly adopted in long-horizon tasks and interact with diverse external environments. Despite their growing capabilities, these agents remain vulnerable to real-world attacks such as indirect prompt injections, which manipulate agents through malicious instructions embedded in external environments. While existing risk assessment work has primarily focused on the injection content, the effect of where an injection appears has not been systematically characterized. Studying injection placement requires agents to operate over longer-horizon tasks and interact with diverse environmental surfaces. Thus, we build three agentic testbeds spanning tool use, coding, and daily workflow domains with diverse injection surfaces. Holding the injection content fixed, we find that agent susceptibility varies across injection placements. We refer to this phenomenon as the trust region effect: agents implicitly assign different levels of trust to different regions of their environments. For example, an agent powered by GPT-5.4 follows an unsafe injection instruction placed in a repository documentation file (i.e., a README or TODO file) 2.6 times more often on average than the same instruction placed in a code or data file. Building upon this observation, we develop learning-based attackers that generate injection content and select better injection placements. Our approach demonstrates that both where an injection is placed and what content it contains can affect the strength of indirect prompt injection attacks. To the best of our knowledge, this work provides the first systematic characterization for the effectiveness of different injection placements, spanning both coarse-grained environmental surfaces and fine-grained positions. The uneven trust that agents implicitly place across their environments provides important insights for future attack and defense design.

cs.AI · cs.CR · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38851

Learning to Reason from Reference Samples Alone: Reinforced Adversarial Learning for Modern LLM Post-Training

Albert Bou, Javier Valdes-Aleman, Jon M. Laurent, Samuel G. Rodriques, Andrew D. White, Michaela Hinks, Siddharth Narayanan

原文摘要

In many settings where LLM post-training is desirable, high-quality answers are recognizable but difficult to score with explicit rewards. Reinforcement learning with verifiable rewards (RLVR) is limited to tasks where an explicit correctness signal is available. Prior adversarial approaches to training reasoning LLMs either complement verifiable rewards with co-trained critics or require reference reasoning traces. We ask whether a co-trained LLM judge can turn reference answers alone into the sole reward for on-policy post-training of reasoning LLMs. We call this method Reinforced Adversarial Learning (RAL). We first validate RAL on three verifiable tasks (integer multiplication, code-based arithmetic-expression synthesis, and SMILES-to-formula translation), using the verifiers only for evaluation. With sufficiently capable judges and enough training, RAL matches direct RLVR. We then apply RAL to scientific abstract generation in two settings. In the first setting, RAL guides the generator to produce abstracts in the scientific domain of the reference abstracts, even though the prompt leaves that domain unspecified. In the second setting, we generate abstracts for U.S. National Institutes of Health grant proposals conditioned on the grant title. Human scientists and an ensemble of LLM judges find our generations harder to distinguish from real abstracts than those of frontier-LLM baselines. These results suggest a path toward post-training reasoning models in scientific and other high-value domains, where increasingly capable judges infer hard-to-specify reward functions from example answers.

cs.AI · cs.LG

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38850

Autonomous Quantum Measurement with Agentic AI

Eunbi Rha, Abhijatmedhi Chotrattanapituk, Mouyang Cheng, Chung-tao Chou, Shenyong Ye, Ashley S. Dale, Long Ju, Luqiao Liu 等 9 位作者

原文摘要

Quantum materials exhibit diverse electronic and functional phenomena, but their characterization requires extensive experimental expertise, iterative decision-making and accumulated laboratory knowledge that is difficult to document and transfer. Here we introduce Quantum Agent (QUANTA), an agentic artificial intelligence (AI) framework for autonomous quantum-materials characterization. Using a widely adopted Physical Property Measurement System (PPMS) as an experimental platform, QUANTA combines retrieval augmentation using scientific literature, experimental protocols and prior measurements with specialized agents for experimental planning, troubleshooting, and data interpretation. A separate deterministic control layer validates instrument actions. Guided by incoming data, QUANTA evaluates measurement quality, revises experimental plans and selects subsequent measurements. We demonstrate these capabilities through transport measurements of the topological Weyl semimetal TaP, as well as investigations of nonlinear responses in magnetic material Co2MnGa. These results establish QUANTA as a generalizable framework for evidence-driven, safeguarded and closed-loop quantum-materials characterization.

cond-mat.mtrl-sci · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38833

Discovering Jailbreaks via Curriculum of Models

Michał Burzyński, Noah T. Rossi, Shomit Basu, Andy Arditi

原文摘要

Automated jailbreak discovery aims to find prompts that elicit prohibited behaviors from safety-trained language models, helping developers identify vulnerabilities before deployment. One class of approaches uses reinforcement learning to train an investigator model to generate prompts that elicit a specified behavior from a target model. However, training directly against a safety-trained target is difficult because nearly all prompts are refused, leaving the investigator with a sparse reward signal. Prior work addresses this with the propensity bound (PRBO), a proxy reward estimated by sampling responses from a more compliant target, and then adjusting for how likely those responses would have been under the original target. We introduce Curriculum of Models (CoM), an alternative method that trains the investigator against a sequence of modified versions of the target, progressing gradually from a version that readily exhibits the specified behavior to the original target. For jailbreak discovery, training begins against a target version that rarely refuses harmful requests, and advances to the next, harder version once the investigator reliably elicits the behavior from the current one. We evaluate CoM on elicitation tasks involving CBRN materials, explosives, and illicit drugs, using two open-weight targets, gemma-4-31b-it and gpt-oss-120b. We compare against PRBO and direct training on the original target without a curriculum. When the investigator is trained to elicit a single behavior, CoM finds successful jailbreaks more reliably than either baseline. When trained on multiple behaviors simultaneously, the CoM-trained investigator can elicit held-out behaviors, though PRBO-trained investigators achieve higher success rates. Our results suggest that a simple curriculum over target models offers a promising approach for addressing sparse rewards in rare-behavior elicitation.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38825

Evaluating Visual Realism of AI-Generated Videos with Eye Tracking

Danqing Shi, Lan Jiang, Jake Lyons, Miri Zilka

原文摘要

Human evaluation of AI-generated videos has so far relied on explicit human feedback, such as judgments or ratings. While these signals capture the conclusions viewers reach, they fail to offer insights into the perception that leads to them. This paper investigates eye tracking as a complementary human evaluation signal for AI-generated videos. We conducted two Visual Turing Test (VTT) experiments with a total of 90 participants on videos generated by latest generative models, in which participants watched each video twice and then gave their feedback. Across the experiments, we collected 4,286 valid eye-tracking trials on 438 videos. Our analysis indicates that gaze patterns offer informative evaluation signals that align with explicit human feedback. Building on the findings, we propose the first computational method GazeEval to predict the perceived visual realism of AI-generated videos from eye-tracking data. It achieves 71.3% accuracy in prediction, providing evidence that eye tracking data carries informative signals about how people evaluate a video. We release the study data to support further research on eye tracking for evaluating AI-generated video.

cs.HC · cs.AI · cs.CV

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38822

CopyLoop: Fast Looped Transformer Decoding via Context Copying and Shallow Screening

Wenyu Du, Stephen Chung

原文摘要

Looped Transformers increase computational capacity through the repeated application of shared blocks, but their recurrent depth makes conventional full-depth decoding expensive. While existing speculative pipelines accelerate generation by processing consecutive token positions at varying depths, they are typically limited by sequential decoding overheads. We introduce CopyLoop, a training-free decoding algorithm that combines low-cost context-copy proposals with shallow screening and continuous recurrent scheduling. Rather than step-by-step generation, CopyLoop proposes multi-token suffixes from repeated context. It then evaluates them early at a shallow recurrent depth to efficiently filter out incorrect branches, processing newer blocks concurrently while earlier blocks are still progressing toward full-depth verification. We evaluate CopyLoop across six model checkpoints from the Ouro, Raven, and Huginn families, covering code generation and multi-step mathematical reasoning. CopyLoop achieves decoding speedups of 2.98x--10.70x over full-depth autoregressive decoding on Base models and 2.74x--3.52x on Thinking models, consistently outperforming or closely matching all competing baselines across evaluation tasks. Furthermore, ablation studies confirm a strong synergy between shallow screening and continuous scheduling, while our analyses show that CopyLoop requires only lightweight model calibration and yields greater speedups on longer generations with reusable contexts.

cs.CL · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2605.08441

DUET: Optimize Token-Budget Allocation for Reinforcement Learning with Verifiable Rewards

Haoyu Hu, Xuandong Zhao, Xuhai "Orson'' Xu, Nori Jacoby

原文摘要

Reinforcement learning with verifiable rewards (RLVR) generates hundreds of thousands of tokens per training step, with rollout generation dominating the computational cost. The overall token budget can be controlled along two main dimensions: (i) deciding which prompts to allocate rollouts to, and (ii) deciding how long each rollout should be. Prior work has generally controlled only one of these dimensions at a time. We show that jointly tuning both decisions under a shared compute budget improves both reasoning quality and wall-clock training time. We instantiate this view as \textbf{DU}al-controlled tok\textbf{E}n alloca\textbf{T}ion (DUET), a computationally efficient layer over GRPO that uses a lightweight pre-rollout surrogate of prompt informativeness to set how many rollouts each prompt receives, and a marker-gated abort rule with importance reweighting to set when to stop them. On Qwen3-1.7B trained on MATH, DUET outperforms full-budget GRPO and the other three budget-aware baseline methods. DUET's advantage further generalizes to other benchmarks across math and coding, and is on par with the best baseline on the scientific Q\&A domain, while also achieving a $1.62\times$ wall-clock speedup. More notably, using only 50\% of the token budget, DUET still outperforms all baseline methods at their full budget, achieving an even higher $2.51\times$ speedup over full-budget GRPO. We verify the high performance of DUET on other backbone LLMs, including Qwen3-4B to 14B and Llama-3.2-3B-Instruct. Notably, the gap between DUET and the strongest baseline \emph{widens} as the budget tightens, contrary to the usual pattern in which efficient methods trade off quality as compute decreases. More broadly, these results suggest that DUET budget-aware control strategies are valuable not only for accelerating training, but also for improving the quality of the learning signal.

cs.LG · cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2609.30177

Search-Aware Reinforcement Learning for Multi-Component Query Understanding in Roblox Game Search

Nayoung Choi, Shengjian Chen, Xiaokai Wei, Wenzheng Zhang, Daiyao Yi, Rachit Pareek, Vincent Su, Michelle Gong 等 9 位作者

原文摘要

Query understanding (QU) in production search systems has typically relied on separate models for each subtask, such as search intent classification, attribute extraction, and query rewriting. Large language models (LLMs) now make it possible to consolidate these subtasks into a single model that emits a unified QU output. This consolidation, however, introduces a new optimization challenge. A single model serves multiple components, each with a distinct downstream role within the search pipeline, and the utility of each depends on how the underlying search engine acts on it rather than on agreement with static labels. We present a search-aware reinforcement learning (RL) framework for multi-component QU based on a distill-then-RL paradigm. Teacher-student supervised fine-tuning (SFT) first yields a schema-compliant small policy. The RL stage then starts from this policy, executes each QU plan on the live production search engine, and optimizes every component with a reward tailored to its operational role, rather than relying on a single reward tied to the final search outcome. On a large-scale user-generated game platform, this component-specific optimization raises NDCG@20 by 8.9 points over the SFT policy and by 3.5 points over a single end-to-end reward. The resulting small, low-latency student models also surpass their much larger teacher, and our final model is being incrementally deployed in production.

cs.AI

查看论文
arXiv ↗PDF ↗
版本更新 · v2arXiv:2609.01202

Towards AI-Assisted Clinical Trial Matching: Practical Considerations, Multicenter Evaluation, and Real-World Deployment

Yin Fang, Qiao Jin, Shubo Tian, Lauren He, Maya Geer, Noor Naffakh, Ryan Huu-Tuan Nguyen, Zifeng Wang 等 27 位作者

原文摘要

Clinical trials are essential for evidence-based medicine and drug development, but many fail because of insufficient patient enrollment. Although there is growing interest in using AI to support patient recruitment, existing systems largely perform eligibility assessment alone and have rarely been evaluated in real-world workflows. Here we present TrialGPT 2.0, an AI-assisted clinical trial recommendation system improved for responsible real-world deployment. Rather than asking only whether a patient may qualify, the system assesses which trials warrant further consideration given the patient's current clinical needs and local workflow priorities, and provides structured, inspectable explanations for expert review. Importantly, we evaluated TrialGPT 2.0 retrospectively and prospectively across government research, patient-advocacy, academic medical center, and community-based radiation-oncology settings. In retrospective multicenter cohorts comprising 288 cases, TrialGPT 2.0 retrieved at least one clinician-recommended trial in its top 10 recommendations for approximately 91% of cases while reducing clinician screening time by 55.0%. In a six-month prospective evaluation embedded in an active precision oncology tumor board, TrialGPT 2.0 contributed additional trial opportunities missed by the routine workflow, increasing the number of patient cases with at least one final trial recommendation by 90.9%. To support scientific reproducibility, we also developed NIH-TrialBench, a clinician-authored benchmark comprising 126 diverse synthetic patient vignettes and matching scenarios from 11 NIH Institutes and Centers. Together, these findings support the value of AI to assist clinical trial matching by improving clinician efficiency and identifying frequently overlooked trial opportunities, ultimately helping to expand and accelerate accrual to cancer trials.

cs.CL · cs.AI · cs.CY

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38755

Leaving, Staying, or Living There? Duration Commonsense Reasoning with Human-LLM Co-annotations

Chenyang Yu, Zhaomin Xiao, Yan Huang, Eduardo Blanco

原文摘要

Gardening in the backyard implies living there, while checking into an Airbnb more likely means staying for a few days. Such duration judgments are important for understanding human mobility, travel behavior, and social context. However, duration is rarely stated explicitly in social media posts and must instead be inferred from noisy textual, visual, and contextual cues. In this paper, we investigate whether vision-language models possess such duration commonsense. We create a new corpus of geotagged tweets co-annotated by humans and LLMs to serve as a testbed for this task. We find that state-of-the-art models struggle with duration inference, even when guided by multi-turn Chain-of-Thought prompting. Error analyses further reveal that models often fail to integrate critical contextual information, or they overly rely on prominent visual details while overlooking subtle hashtags and semantic nuances. These findings underscore the significant challenges remaining in context-aware multimodal understanding.

cs.AI · cs.IR

查看论文
arXiv ↗PDF ↗
新论文 · v1arXiv:2610.38744

AtomTree: A Hierarchical Framework for State-Aware Embodied Instruction Following with LLMs

Haotian Zhai, Tianming Sha, Junnan Li

原文摘要

Embodied instruction following in interactive environments requires agents to interpret natural language commands and complete long-horizon tasks. Existing planners either rely on large labeled trajectories, or use few-shot and zero-shot LLM prompting that produces a whole plan in a single serial reasoning process. These approaches depend on linguistic common sense, commit to a plan from the initial instruction, and only update after failures, which weakens robustness and state awareness. We present AtomTree, a hierarchical framework for state-aware task planning with LLMs. AtomTree incrementally builds a tree of subgoals and actions, where the LLM decides at each step, from the current symbolic state, whether to further decompose or to output a grounded low-level action. A lightweight memory module summarizes past outcomes for reuse in later decisions. When a low-level action stalls, AtomTree triggers hierarchical backtracking replanning to return to the parent node and regenerate the subtree plan from the latest state, preventing continued drift along an incorrect trajectory. In zero-shot evaluation on the ALFRED benchmark, AtomTree outperforms prior LLM-based planners. Project page: https://zhaihaotian.github.io/atomtree-website/.

cs.RO · cs.AI · cs.CL

查看论文
arXiv ↗PDF ↗