The study examines how four large language models (Gemma-3-4b-it, Qwen2.5-7B-Instruct, Llama-3.1-8B-Instruct, and GPT-5-nano) exhibit repetition effects in a social media simulation. Using a two‑phase within‑context design, researchers collected 336,000 ratings on truth, importance, sentiment, and interest for 100 statements across 10 feed variants and 3 replications. Results show distinct patterns: Gemma-3 displays a genuine Illusory Truth Effect, Qwen2.5 shows a mere exposure effect, GPT-5-nano shows no truth boost and mild skepticism, and Llama-3.1 shows a small truth boost with reduced evaluative dimensions.
By Azza Bouleimen, Nicol\`o Pagan, Anik\'o Hann\'ak
The paper describes progress toward an AI Software Factory that accelerates every stage of the software development lifecycle—Targeting, Coding, Reviewing, and Ops—by generating a metadata exhaust for self‑improvement. It focuses on data systems and evolutionary coding tasks, reporting scaled deployments at Microsoft that achieved three‑fold engineering efficiency over agentic coding and up to 22‑fold token efficiency. The authors also outline several open challenges for further development.
By Anna Pavlenko, Bogdan Crivat, Brandon Haynes, Carlo Curino, Fotis Psallidas, Jaro Slawinski, Johannes Freischuetz, Laura Pereira Sanchez, Markus Weimer, Mathieu Demarne, Matthias Jasny, Mauktik Gandhi, Max Bovykin, Mirco Milletari, Purbasha Ghosh, Qiushi Bai, Raghu Ramakrishnan, Rahul Pandita, Sergiy Matusevich, Shivaram Venkataraman, Subru Krishnan, Md. Tareq Mahmood, Tiemo Bang, Venkatesh Emani, Xuan Zhao, Yiwen Zhu
ThuRunel is an AI advisory agent designed for high‑stakes domains that exhibit a two‑phase interaction structure: an initial empathetic elicitation phase followed by a specialist judgment phase. It uses a finite‑state belief management framework, a chain‑of‑thought teacher synthesis protocol, and learned generation adapters to dynamically decide what to ask, when to stop, what to resolve autonomously, and what to forward to a specialist. In evaluations against eleven baselines, ThuRunel consistently improves elicitation completeness and the quality of specialist briefs, and it is publicly deployed as a bilingual web application that cites its knowledge base in every answer.
By Yuyan Chen
The paper introduces PhenoAIR, a reliability‑aware multi‑agent framework for predicting mechanisms of action (MOA) from Cell Painting morphological profiles. It treats retrieved neighbors as uncertain evidence, calibrating their reliability based on source, phenotype stability, and mechanism confusion, and refines predictions through controller‑guided evidence evaluation. PhenoAIR is benchmarked on JUMP Cell Painting data and outperforms both representation‑matching and LLM‑based baselines across controlled, realistic, and open‑world settings.
By Jiayuan Chen, Botao Yu, Tianyu Liu, Thai-Hoang Pham, Meng Wu, Ping Zhang
MAADBench is a refreshable benchmark for anomaly detection in multi‑agent systems powered by large language models. It addresses the challenge of keeping benchmarks current by sampling and coupling generative tasks, generating trace data under configurable LLM backbones, and automatically providing deterministic step‑level labels. The authors evaluated 25 anomaly‑detection methods on 5,200 labeled traces, finding that existing approaches depend heavily on supervision, struggle with subtle MAS‑specific anomalies, and lack robustness across different LLM backbones.
By Lei Ma, Dennis Hofmann, Haowen Xu, Joshua DeOliveira, Peter VanNostrand, Lei Cao, Elke Rundensteiner
The paper introduces BRIDGE, a bilevel optimization framework that jointly trains a large language model (LLM) and a retriever for agentic reinforcement learning (ARL). It demonstrates that adapting the retriever before the policy yields better rewards, and that BRIDGE outperforms existing methods on seven open‑domain QA benchmarks and medical QA tasks, achieving significant gains in accuracy and reasoning quality.
By Quan Xiao, Mingda Liu, Gaowen Liu, Katsuki Fujisawa, Tianyi Chen
The paper introduces TrustSwap, a counterfactual test that swaps or removes source reliability labels while keeping evidence text constant, to evaluate how retrieval‑augmented fact‑checking models respond across verdict, confidence, and search decisions. Experiments on untrained and RL‑trained models show that confidence and search largely follow labels, yet label changes can flip a significant portion of verdicts, especially in larger models. The authors propose trust‑swap augmentation (TSA) to mitigate this shortcut, demonstrating reduced verdict flip rates and maintained accuracy in several settings, though its effectiveness diminishes at larger model scales.
By Jianchang Su, Yiwei Yang, Wei Zhang
The paper introduces Agent Distillation, a framework for transferring task‑solving knowledge from a teacher agent to a student agent. It categorizes where this knowledge is retained—within the model, as artifacts, through the execution harness, or across substrates—distinguishing transfer evidence from outcomes. An evaluation framework is proposed to link retention to causal contribution and practical utility, aiming to support reliable, maintainable, and safe agent development.
By Ziluowen Luo, Senzhang Wang, Chaozhuo Li, Jun Yin, Hao Yan, Ming Cheng, Chenxu Wang, Songyang Liu, Litian Zhang, Qiwei Ye, Zheng Liu, Philip S. Yu
The paper introduces Synthetic Universes, a benchmark that pairs well-known theoretical worlds with closely related twisted variants governed by noncanonical mechanisms. It evaluates scientific agents on each law twice: by testing predictive performance on unseen data and by checking if the law recovers the underlying generating mechanism. Results from a 60‑cell study show a dissociation between predictive adequacy and mechanism recovery, indicating that these are distinct scientific claims requiring separate tests.
By Kargi Chauhan
The paper introduces LibraryDesignBench, a benchmark that tests how well agents can design reusable libraries from specifications without prescribed designs. It evaluates libraries by measuring the correctness and simplicity of programs written by three different user agents across 242 programming problems in four languages. Findings show that while agents often replicate human-designed abstractions, downstream agents still tend to reimplement library features due to rigidity or usability issues, and that providing more prescriptive guidance improves reuse and program simplicity.
By Gabriel Orlanski, Alex L. Zhang, Avi Trost, Vincent Sunn Chen, Frederic Sala, Aws Albarghouthi, Ludwig Schmidt
The paper introduces the Distinguish-or-Homogenize principle, where an agent can either spend resources to differentiate between latent fault models or alter the system state so that the remaining models share a common acceptable policy, eliminating further diagnosis. This leads to the Last-Chance Policy Identification (LCPI) framework, which evaluates correctness at the reached state rather than the initial one, and defines the Last Identifiable Margin (LIM) as the boundary between distinguishing and homogenizing. For deterministic diagnostic graphs, an Exact-LIM recursion is provided, while for noisy finite-horizon recovery the authors propose Risk-Budgeted Compatibility Planning (RBCP), which searches a compatibility-aware frontier under a hard worst-case failure constraint, demonstrating improved risk-feasible recovery in microservice and MiniGrid scenarios.
By Yibo Guo, Xiaodan Wang
Code4Scene is a benchmark that evaluates coding agents on constructing and editing 3D scenes in Unreal Engine. It tests agents on two tasks: construction, where they must build a scene from open‑ended language, and editing, where they must recover a target scene from reference images while preserving everything else. The benchmark measures task fulfillment, artifact integrity, and physical validity, revealing that construction and editing performance are correlated but not interchangeable, with agents struggling most with spatial composition and precise edits.
By Xiaokang Ye, Siddhant Hitesh Mantri, Zimeng Chen, Edward Zhang, Zhaoxu Zheng, Yuanheng Li, Yizhao Chen, Tianyang Huang, Lianhui Qin
The paper explores using large language models (LLMs) to replace traditional compiler backends, a process termed AI lowering. An LLM agent translates Triton kernels directly into NVIDIA PTX, achieving 0.83x–3.34x the performance of autotuned Triton on a variety of GPUs and ML kernels. The study also extends a PTX verifier to support modern GPU features, highlighting the potential for AI compilers to reduce engineering effort for new hardware.
By Fran\c{c}ois Costa, Charly Castes, Thomas Bourgeat, Azalia Mirhoseini
The paper investigates how large language model agents that use tools respond to changes in plan priorities versus default plan removal, a phenomenon termed the "default trap." Experiments across 3,200 decision windows on Retail, Airline, and AgentDojo tasks show that switching priorities strongly redirects model choices, while removing a default plan yields weaker responsiveness. Additional studies reveal that the order of account lists and the presence of extra text significantly influence default target selection and priority effects, with overall task success varying from -19.4 to +8.3 points relative to no plan.
By Xueqi Li, Jingjie Ning, Yibo Kong
The paper introduces neuro‑symbolic computer use, a method that learns reusable policies to execute recurring computer workflows efficiently. Instead of re‑planning each run, the learned policy encodes stable decisions (ordering, variables, loops, branches) into executable code while delegating observation‑dependent decisions to neural models. Using neuro‑symbolic policy iteration, the approach iteratively refines the policy from a single agent trajectory, diagnoses failures, and revises the code with a coding model, achieving superior Pass^3 scores and significant reductions in per‑run cost and latency on OSWorld‑Verified and ScienceBoard benchmarks.
By Hyewon Suh, Thanh Minh Nguyen, Chih-Lun Lee, Darrow Hartman, Lizhao Liu, Xin Eric Wang, Ang Li, Jiachen Yang
The paper introduces a Physics‑Informed Multi‑Agent Coordination framework that embeds calibrated BCMP queueing topologies into a decentralized multi‑agent reinforcement learning system for hospital patient flow. It formulates the problem as a Decentralized Partially Observable Markov Decision Process with coupled resource constraints, enabling departmental agents to negotiate patient routing and service scaling while exchanging localized action fingerprints to handle non‑stationarity. Empirical tests on MIMIC‑IV data show the approach reduces cumulative system delay compared to static Markovian models, heuristic dispatching, and independent multi‑agent baselines, all while preserving clinical safety constraints.
By Guoqing Zhang, Rafik Hadfi, Takayuki Ito
MatToolBench is a new benchmark that evaluates multimodal GUI agents on professional materials science software. It contains 204 tasks across 10 tools in three modalities—GUI operation, OriginPro scripting, and code-based database queries—executed inside a Windows 11 VM. The benchmark offers fine-grained, expert-decomposed scoring and a high-performing multimodal judge for aesthetic assessment, revealing that strong general benchmark performance does not transfer to scientific workflows.
By Mei Wu, Rui Xie, Runyu Zhang, Yuqiang Li, Tianfan Fu, Bo Chen, Kai Yu, Xin Chen, Lu Chen
SkillCome introduces a skill-evolution framework that uses group contrast optimization and a dual memory system to refine large language model skills. By generating multiple trajectories per question and contrasting successful versus failed ones, it identifies key behavioral divergences to guide skill edits. The dual memory accumulates historical evidence across groups, enabling more generalized and reliable optimization signals, leading to consistent performance gains on diverse benchmarks.
By Haolin Li, Feng Hong, Ang Li, Chilin Fu, Weichang Wu, Ya Zhang, Yanfeng Wang, Xiaolu Zhang, Jiangchao Yao
The paper introduces Milestone Viability Potential Policy Optimization (MVPO), a reinforcement learning algorithm designed for long‑horizon large language model agents. MVPO addresses the zero‑credit failure problem by learning from viable failure prefixes, estimating prefix potential over Union‑Find viability regions, and repairing zero‑credit groups with potential‑difference advantages. Experiments on Qwen2.5‑1.5B‑Instruct demonstrate that MVPO outperforms eight strong baselines, improving success rates on ALFWorld and WebShop with minimal overhead.
By Qi Zhou, Yuanfan Li
The paper introduces ToxicBench, a benchmark designed to evaluate how tool‑augmented data agents handle incorrect tool outputs. By pairing clean and poisoned observations across numerical, label, schema, and retrieval errors, the authors assess both the checking process and the final answer adoption. In a 118‑task GPT evaluation, poisoning reduces task success by 26–39 percentage points, revealing that repeated poisoning leads to wrong-answer adoption even after checking, while ordinary retries help only under one‑shot poisoning. Human annotations on 200 trajectories confirm the scoring system’s reliability, showing 96% agreement with task success and supporting the benefits of retries and audit‑based adoption.
By Zifu Tao, Changqing Yin