The paper introduces HMS‑SCP, a hierarchical multi‑scale semantic‑aware cooperative perception framework for V2X communication. It uses a spatial importance predictor to select task‑relevant grid elements at multiple scales and maps them directly into complex‑valued symbols for joint source‑channel coding, achieving ultra‑low symbol rates and noise resilience. Experiments on OPV2V and DAIR‑V2X show that HMS‑SCP maintains high‑confidence far‑field detection with sub‑16 ms latency even under severe Rayleigh fading and extreme compression.
By Chun-Yeow Yeoh, Chee Keong Tan, Joanne Mun-Yee Lim, Heng-Siong Lim
The paper introduces policy‑counted DecPOMDPs, a variant of decentralized partially observable Markov decision processes that mitigates the exponential growth in agent numbers by counting policies instead of agents. By exploiting symmetry among agents, the authors achieve a compact representation that reduces model complexity and evaluation cost to polynomial levels. They further present a dynamic programming algorithm that leverages this compact form to solve policy‑counted DecPOMDPs efficiently.
By Nazl{\i} Nur Karabulut, tanya Braun
The paper evaluates out‑of‑the‑box object detection models for automatic target detection and recognition (ATD/R) in military settings. Six YOLO variants and two DETR variants were benchmarked on a new military dataset featuring vehicles, occlusions, and small targets, with performance measured in mAP@0.5 and mAP@0.5:0.95 across air‑to‑ground and ground‑to‑ground perspectives. Findings show larger models and DETR-based approaches perform best, fine‑tuning on the VisDrone dataset improves air‑to‑ground and small‑object performance, yet all models still struggle with small targets in air‑to‑ground scenarios.
By Alma M. Liezenga, Lotte Nijskens, Henrik R. Baumann, Stefan Becker, Simon Bensberg, Niccol\`o Camarlinghi, H{\aa}vard R. Eiring, Alexander W. Johnsgaard, Tanel Liiv, Giuseppe Martino, Matteo Marturini, Matthias Rapp, Jan Erik van Woerden, Alexander Wolpert, Hugo J. Kuijf
The paper surveys the growing use of AI agents that modify external state via the Model Context Protocol (MCP) ecosystem, noting an increase from 27% to 65% of tool use. It argues that when such agents operate on public blockchains, the blockchain execution layer’s properties—irreversibility, signing authority, continuous autonomy, and sequence-level composition—reshape the threat model, making failures irreversible. The authors organize existing MCP-security literature into an attack-surface taxonomy, present a Web3 risk-mapping matrix linking attack classes to amplified impacts and mitigations, and conclude that current defenses are inadequate, stopping fewer than 30% of attacks and less than 3% of model-level safety failures.
whyItMatters":"The study highlights that AI agents acting on Web3 introduce irreversible risks that conventional software security cannot address, underscoring the need for stronger, blockchain-aware safeguards."
By Rabimba Karanjai (Larry), Yang Lu (Larry), Nour Diallo (Larry), Wujie Xiong (Larry), Lei Xu (Larry), Weidong (Larry), Shi
The paper introduces eXplainable DFT (XDFT), a self‑evolving computational agent that transforms experiment‑simulation mismatches into executable searches for physical mechanisms. XDFT formalizes candidate mechanisms as hypotheses, tests them against experimental data, and refines its search strategy through a learning loop. In a benchmark of 112 cases where standard calculations predicted a metal but experiments found a semiconductor, XDFT resolved 105 cases with evidence‑supported mechanisms, and its top‑ranked hypotheses improved dramatically over initial expert priors.
By Yue Li, Penghui Yang, Yushan Xiao, Zhonghan Zhang, Jianguo Huang, Yuhao Lu, Cuntai Guan, Bo An, Bijun Tang, Zheng Liu
The paper argues that cross‑view correspondence, commonly used in agent evaluation and trace‑based learning, functions as a measurement intervention. Removing or altering this correspondence can create artificial sensitivity or invariance, and multiple optimal correspondences can obscure mechanism labels and learning credit. The authors propose a validity theory with two‑sided validation, all‑optima identification, and uncertainty propagation, and demonstrate through experiments that unvalidated correspondences can misattribute credit and erase harmful responses.
By Zhen Zhang, Ahmad Hafez, Amr Alanwar
SOD: Step-wise On-policy Distillation for Small Language Model Agents proposes a new framework that adaptively reweights distillation strength at each reasoning step based on step-level divergence. This approach mitigates cascading errors in tool-integrated reasoning by attenuating misleading teacher signals in high-divergence regions while preserving dense guidance where student and teacher align. Experiments on math, science, and code benchmarks show up to 20.86% improvement over the second-best baseline, with a 0.6B student scoring 26.13% on AIME 2025.
By Qiyong Zhong, Mao Zheng, Mingyang Song, Xin Lin, Jie Sun, Houcheng Jiang, Xiang Wang, Junfeng Fang
The paper shows that fine‑tuning a large language model (LLM) with a debate framework—where a generator and a critic compete and a weaker LLM judge adjudicates—reduces reward hacking compared to standard reinforcement learning from AI feedback (RLAIF). In experiments on mathematics tasks, the debate approach keeps the judge’s performance stable, achieving a 45% higher peak validation accuracy than the RLAIF baseline and mitigating the rapid exploitation of judge errors. Additional findings indicate that weakening the judge speeds hacking unless countered by extra debate rounds, that debate can override misalignment prompts, and that word‑limit constraints on critiques help balance the game and prevent judge hacking.
whyItMatters":"The study demonstrates a practical method to curb reward hacking in RL‑based AI systems, addressing a key obstacle for safely scaling AI oversight."
By Zachary Kenton, Lili Janzer, Rory Greig, Tian Huey Teh, Kirill Tyshchuk, Jonah Brown-Cohen, Harri Edwards, Senthooran Rajamanoharan, Noah Y. Siegel, Natasha Jaques, Rohin Shah
StagedWorkspace is a versioned workspace designed for knowledge‑work AI agents, ensuring that every parsed view, native file edit, and review diff is explicitly tied to a specific version of the workspace state. By binding parsed records and review diffs to content hashes of native files, the system improves performance on tasks such as OfficeQA and APEX‑Agents, achieving higher pass rates and rubric scores compared to single‑view approaches. The study demonstrates that providing dual parsed/native access and visible diffs enhances agent performance, highlighting workspace state as a key experimental variable for future benchmarks.
By Yining Hua, Hongbin Na, Yifan Zhou, Akshay Kalose, Cyrus Ayubcha, Levi Lian
The study examines how users of a major dating platform respond to autonomous LLM agents that converse on their behalf. Using two large surveys, researchers built a latent-variable model showing that willingness to send and receive agent-mediated messages are highly correlated yet distinct. The findings reveal a delegation asymmetry: users are more willing to deploy their own agent than to engage with others’ agents, leading to low overall reciprocity and gender‑directional imbalances in agent interactions.
By Daria Leshchikova, Valentina V. Kuskova, Dmitry Zaytsev, Valerii Klimov
The paper introduces a mesh‑free policy iteration framework that blends classical dynamic programming with physics‑informed neural networks (PINNs) to solve high‑dimensional, nonconvex Hamilton–Jacobi–Isaacs (HJI) equations. The method alternates between solving linear second‑order PDEs under fixed feedback policies and updating controls via pointwise minimax optimization using automatic differentiation. The authors prove local uniform convergence of the value function iterates to the unique viscosity solution under standard Lipschitz and uniform ellipticity assumptions, and demonstrate the approach’s accuracy and scalability in two‑, five‑, and ten‑dimensional stochastic games, outperforming direct PINN solvers.
By Hee Jun Yang, Minjung Gim, Yeoneung Kim
The paper proposes EvalXRL, a benchmark that evaluates explainable reinforcement learning (XRL) methods by measuring how well their explanations help a large language model coding agent diagnose and fix bugs in RL agents. Unlike current metrics that focus on faithfulness or human ratings, EvalXRL uses a closed‑loop, scientific‑method style interaction where the coding agent repeatedly invokes XRL methods, refines hypotheses, and repairs the agent, scoring each method by the resulting RL reward. This approach enables a head‑to‑head comparison of multiple XRL techniques in realistic debugging scenarios.
By Ram Rachum, Yotam Amitai, B\'alint Gyevn\'ar, Reuth Mirsky, Cameron Allen
The paper presents dynamic shields for AI-controlled autonomous systems, enabling runtime safety enforcement that adapts to changing safety specifications without recomputing from scratch. Unlike traditional static shields, these dynamic shields are pre-designed for a set of possible safety parameters and can quickly adjust as the true specification becomes known during operation. Experiments on robot navigation in unknown terrains show that dynamic shields require only a few minutes offline and a fraction of a second to a few seconds online, outperforming brute-force recomputation by up to five times.
By Davide Corsi, Kaushik Mallik, Andoni Rodriguez, Cesar Sanchez
The paper investigates a Multi-Scale Spectral Attention Module (MSAM) for hyperspectral image segmentation in autonomous driving. MSAM uses three parallel 1D convolutions with different kernel sizes (1–11) and adaptive feature aggregation, integrated into UNet’s skip connections. Experiments on urban driving datasets show that MSAM improves mIoU by 2.32% and mF1 by 2.88% over baseline UNet-SC while keeping GPU performance competitive, with optimal kernel combinations varying by dataset.
By Imad Ali Shah, Jiarong Li, Tim Brophy, Martin Glavin, Edward Jones, Enda Ward, Brian Deegan
StartupBench is a new benchmark that evaluates general-purpose agents on end-to-end workflows derived from AI startup products that have proven market adoption. It translates real-world product workflows into deliverable-oriented tasks and assesses them with detailed rubrics. The study finds that even the best models complete only about 30% of these tasks, highlighting challenges such as complex instruction following and domain expertise.
By Liya Zhu, Xin Ma, Tao Liu, Haodong Wang, Ge Zhang, Jingzhe Ding, Qingshui Gu, Yongjie Zhong, Jinxiang Meng, Yuan Gao, Yunqiu Zhou, Hao Zhu, Jifeng He, Yongzhi Liao, Xinyi Zhang, Chaoxin Li, Yi Zhu, Xi Lin, Duju Zeng, Xiang Gao, Wen Zhang, Yunyang Wang, Duo Wang, Huan Zhou, Zuo Wang, Jin Chen, Kaiyuan Zhang, Chuqian Yu, Tianhao Yu, Longxiang Liu, Jianbo Xue, Huimin Che, Jiahao Wang, Yujia Qin, Jiaheng Liu, Shen Yan, Xiaolong Chang, Wenhao Huang
The paper re‑evaluates memory‑based self‑improving agents by adding multiple runs to measure variance and by randomizing task order. It finds that agent performance is noisy in complex, multi‑step environments and that improvement depends heavily on the sequence of tasks, revealing a hidden curriculum effect. The authors suggest that underspecification of tasks and environments contributes to this fragility and demonstrate that adding detailed rubrics and feedback can partially mitigate performance drops, though gaps remain.
By Qinyuan Ye, Yu Li, Yada Pruksachatkun, Jiaxin Zhang, Chien-Sheng Wu
AutoResearch is a two‑stage autonomous research system that links Idea Generation with Idea Execution. In the generation phase it blends new research signals with existing domain knowledge, identifies transferable mechanistic insights, and produces grounded, testable research plans through multi‑model generation and cross‑review. The execution phase then decomposes these plans into experiments, iteratively implements and diagnoses them, and uses independent evidence‑based review to accept or revise conclusions, thereby turning ideas into measurable progress while minimizing hallucinations.
By Yiming Ren, Xiang Liu, Qumeng Sun, Xiao Zhang, Jiahao Li, Haoyang Zhang, Junjie Wang
Teach-and-Grow Learning (TGL) is an agent-centered architecture that transforms a few successful demonstrations into reusable Skill Blocks, enabling a robot to compose, execute, and revise behaviors in new scenes without task-specific policy retraining. The system maintains a Skill Library and structured Experience Memory to capture successes, failures, and repairs, allowing persistent reuse and agent-directed adaptation. Evaluation on the LIBERO benchmark shows state-of-the-art performance, and the authors propose a scaling-law hypothesis suggesting that accumulated reusable experience reduces future-task error and teaching demand following a power-law trend.
By Chang Nie, Zhe Liu, Hesheng Wang
PathPocket is a multimodal AI co‑pilot that grounds pathology decision‑making in evidence. It builds the largest pathology evidence corpus (≈110,472 documents) and a hypergraph of 4.55 million entities and 7.10 million relations to support traceable reasoning. The system handles text and multimodal queries, including ROI and gigapixel whole‑slide images, and outperforms current state‑of‑the‑art models on a benchmark of over 200,000 real‑world cases, improving pathologists’ diagnostic accuracy and confidence.
By Zhe Xu, Zhengyu Zhang, Zhiyuan Cai, Jiahao Xu, Yijie Lin, Ziyi Liu, Junlin Hou, Hongyi Wang, Yuxiang Nie, Yihui Wang, Jiabo Ma, Ling Liang, Yingxue Xu, Zhengrui Guo, Guanghao Wu, Danyi Li, Ziqi Zhou, Donglin Tan, Zhijian Cen, Ying Tan, Xiaolin Liu, Qi Xie, Xiaoying Tang, Xi Peng, Cheng Deng, Lijuan Qu, Ronald Cheong Kin Chan, Li Liang, Hao Chen
Agentic ESOpt proposes using evolution strategies (ES) instead of reinforcement learning to fine‑tune large language‑model agents for long‑horizon tasks. ES offers model scalability, flexibility, and better long‑horizon credit assignment, enabling full‑parameter optimization with minimal GPU memory. The framework samples parameter perturbations, evaluates agents with rewards, and updates online, achieving notable performance gains on WebArena‑Lite and in test‑time prompt‑parameter co‑evolution.
By Zhi Zheng, Rongsheng Chen, Yunpeng Ba, Zhenkun Wang, Yee Whye Teh, Wee Sun Lee