MineExplorer is a benchmark designed to assess the open‑world exploration abilities of multimodal large language models (MLLMs) in Minecraft. It filters out tasks that rely heavily on Minecraft‑specific knowledge, organizes tasks into ReAct‑style capabilities, and composes atomic tasks into implicit multi‑hop challenges. A multi‑agent synthesis workflow creates reliable task graphs, sandbox scenes, and rule‑based milestone evaluators, and human evaluation confirms its superiority over a single‑agent baseline. Experiments show that while advanced MLLMs can handle many single‑hop tasks, they struggle with longer trajectories that require coordinating hidden prerequisites, and larger models or different thinking modes do not consistently improve performance.
By Tianjie Ju, Yueqing Sun, Zheng Wu, Wei Zhang, Yaqi Huo, Xi Su, Qi Gu, Xunliang Cai, Gongshen Liu, Zhuosheng Zhang
The paper presents a pipeline that uses Large Language Models (LLMs) to extract information from 536 peer‑reviewed agent‑based modeling papers for systematic literature reviews (SLRs). GPT‑4.1 achieves about 77.95% paper‑level accuracy, while GPT‑5.0 reaches 81.67%. Field‑level accuracy varies widely, and the study notes that agreement between LLMs can signal output quality, with low agreement indicating hallucinations and high agreement with low accuracy suggesting noise in the human dataset.
By Orhan Yagizer Cinar, Timur Emre Ozkose, Emma Von Hoene, Amira Roess, Taylor Anderson, Hamdi Kavak
Nomad is an autonomous system designed to explore and discover insights within large data corpora. It builds an explicit Exploration Map to systematically traverse a domain, generating and testing hypotheses with an explorer agent that leverages document, web, and database searches. After verification, it produces cited reports and meta-reports, and its evaluation framework assesses trustworthiness, quality, and diversity, showing superior performance over baselines on UN, WHO, and arXiv datasets.
By Bokang Jia, Samta Kamboj, Satheesh Katipomu, Seung Hun Han, Neha Sengupta, Andrew Jackson
The article reviews how Large Models (LMs) based on Transformer architectures and self‑supervised pre‑training can address longstanding challenges in Battery Prognostics and Health Management (BPHM). It surveys LM applications across data scarcity, generalization, interpretability, and system automation, and outlines a roadmap for future research, including collaborative data ecosystems, validation, trustworthiness, and efficient deployment. The review aims to guide researchers and practitioners in developing next‑generation battery management systems that are safe, reliable, and autonomous throughout battery lifecycles.
By Jiale Liu, Huan Wang, Weicheng Wang, Rong Zhu, Qiqi Wang, Min Xie
The article discusses how large language models (LLMs) are transforming electronic design automation (EDA) by moving beyond isolated task assistance to a hierarchical framework of roles: Generator, Agent, and Orchestrator. It highlights that current LLM-based solutions often produce plausible but not physically correct hardware, suffer from fragmented tools, and lose design context, which hampers scalability to industrial designs. The authors argue for a standardized, physics-aware Orchestrator that integrates tools and agents across the EDA flow to improve reliability and accessibility of hardware design.
By Matthew Youngman, Cristian Sestito, Themis Prodromakis
The paper introduces SARA, a framework that separates action induction from runtime authorization in tool‑augmented LLM agents. By treating these as distinct roles, SARA uses an Action Probe to record action provenance and only authorizes tool calls that align with the user objective and past successful executions. Experiments on AgentDojo and AgentDyn show that SARA reduces action‑to‑side‑effect risk to below 0.63% while preserving task performance.
By Xiaokun Guo, Zhen Xu, Dongdong Huo, Yanqiu Zhang, Wei Wang, Qinfu Yang, Dongjin Yu, Yu Wang
The paper introduces a new approach to safety in contextual bandits with continuous actions, focusing on high‑probability constraints on the realized cost rather than expected cost. It presents the High‑Probability Constrained UCB algorithm, which balances reward exploration with conservative safety estimation, and provides theoretical regret guarantees for linear models and extensions to general function classes. Experiments demonstrate that this realized‑cost safety framework significantly reduces safety violations compared to expected‑cost constrained methods.
By Spyros Dragazis, Aldo Pacchiano
MemToC is a controlled benchmark that tests how large language models resolve conflicts between their internal memory and tool outputs. It contains 6,504 episodes built from 542 factual questions, each paired with a model‑generated closed‑book answer and a tool return whose correctness is known, creating four distinct source‑correctness scenarios. Across five 7‑9B open‑weight models, tool responses overwhelmingly dominate closed‑book answers, and only a minority of instruction‑tuned models correctly retain a verified answer when the tool is wrong, while most follow a correct tool or repeat a wrong tool.
By Arseniy Varlamov, Rishat Zinnatullin, Elisei Rykov, Alexander Panchenko, Ilseyar Alimova
DIANOIA introduces a diagnostic framework for multi‑agent large language model systems, decomposing reasoning gain into three measurable channels—coverage, fidelity, and synthesis. The protocol identifies bottleneck channels for a given task and implements a corresponding multi‑agent system with role‑diverse proposers, execution‑grounded verification, and iterative synthesis. Experiments on GSM8K, AIME‑2025, MBPP, and BFCL‑SP show that DIANOIA outperforms strong baselines, achieving significant token savings and accuracy gains while accurately pinpointing the critical channels.
By Yiming Yang, Zhuoyuan Li, Fanxiang Zeng, Hao Fu, Yue Liu
PICasso is an AI‑enabled framework that converts natural‑language specifications into manufacturable silicon photonic integrated circuits (PICs) through a structured pipeline of NL → YAML → GDS, PDK‑aware knowledge injection, automated placement and routing, DRC/LVS validation, and SAX‑based photonic simulation. The authors introduce PIC‑Set, a benchmark of 36 parameterized PIC design tasks, and evaluate several large language models (LLMs) using new metrics such as structural and functional Spec@k, optimization efficiency, and robustness. Across the benchmark, PICasso markedly improves specification satisfaction, achieving up to 92.7% structural Spec@3 and 52% functional Spec@3, while reducing mean insertion loss from 4.98 dB to 3.25 dB through simulation‑guided optimization.
By Deepak Vungarala, Deniz Najafi, Abdulrahman Aljoudi, Zahra Ghanaatian, Navid Khoshavi, Gourav Datta, Arman Roohi, Mahdi Nikdast, Shaahin Angizi
GRAIN is a single-agent reinforcement learning framework that improves large language models’ robustness to real‑world shifts in node identifiers and task formulations by treating reasoning as a semantic parsing and tool‑execution pipeline. It introduces a Structure Invariance Reward that validates intermediate graphs against ground‑truth topologies, encouraging the model to learn genuine text‑to‑structure mappings instead of overfitting to surface patterns. On the new GRIT benchmark, GRAIN surpasses multi‑agent baselines by 16.45% in accuracy, reduces latency by about 24%, and halves the out‑of‑distribution gap of fine‑tuned models while remaining robust on large‑scale graphs beyond the training distribution.
By Zike Yuan, Han Zhang, Jianzhi Yan, Le Liu, Cai Ke, Huozhi Zhou, Jian Xie, Jiran Yin, Yukun Cao, Yue Yu, Hui Wang, Ming Liu, Bing Qin
The paper investigates how changes in prompts and persona names affect the behavior of generative agents in an epidemic simulation. It finds that synonymous prompts produce negligible differences, while minor prompt variations and contextual changes do influence outcomes. Persona names, even when imbued with distinct identities, do not significantly alter epidemic results.
By Ross Williams, Niyousha Hosseinichimeh
Daydreaming is an execution‑only attack that steals multi‑file agent skills by interacting with a black‑box task service. By adaptively crafting tasks and analyzing the returned results, the attacker reconstructs the hidden skill without ever requesting or revealing it. In experiments on seven skills and four victim models, Daydreaming recovers 86.8% of the original capability using only 32 victim calls on average, outperforming prior methods and demonstrating that hiding skill files and filtering direct disclosure are insufficient defenses.
By Yu-Lin Tsai, Yu-An Lu, Ci-Yang Tsai, Muxi Lyu, Raluca Ada Popa, Chia-Mu Yu
DeepPlanner is an end-to-end reinforcement learning framework designed to enhance the planning capabilities of deep research agents. It introduces an entropy-based advantage shaping mechanism that allocates larger updates to high-entropy planning tokens and selectively upweights sample-level advantages during planning-intensive rollouts. Experiments on seven deep research benchmarks show that DeepPlanner improves planning quality and achieves state‑of‑the‑art results with a lower training budget.
By Wei Fan, Wenlin Yao, Zheng Li, Feng Yao, Xin Liu, Liang Qiu, Qingyu Yin, Yangqiu Song, Bing Yin
AgentJudgeBench is a new benchmark that evaluates the reliability of large language model (LLM) judges on agentic tool‑calling tasks involving workflow directed acyclic graphs (DAGs). It contains 3,808 instances across six DAG topologies and three difficulty tiers, tested with five generators (3B–70B open‑weight models and GPT‑5.4) and six judges (20B to frontier scale) under both paired‑with‑and‑without‑ground‑truth conditions. The study finds that judge alignment degrades with task difficulty, ground‑truth exposure can sometimes hurt alignment, and structured evaluation rubrics provide modest improvements, revealing a structural ceiling that model capacity alone cannot surpass.
By Abhigya Verma, Amit Kumar Saha, Seganrasan Subramanian, Sai Harshitha Aluru
The paper extends Co‑Scientist, a Gemini‑based multi‑agent system, and validates it in real‑world scientific settings. In materials science it designed a safe precursor route for MXenes and achieved single‑attempt growth of monolayer MoS₂, MoSe₂, and WS₂. In biology it predicted swarming phenotypes of engineered E. coli, and in computer science it discovered a superior inference‑time scaling architecture for HealthBench. A double‑blind study with 30 experts showed that Co‑Scientist’s reliability modules reduce hallucination and plagiarism while improving research safety.
By Samuel Schmidgall, Xiaokai Zhu, Marian Shaw, Lin Yang, Valentin Li\'{e}vin, Jingyun Yang, Yuchen Zhuang, Tim Strother, Alex Bijamov, Min Woo Sun, Anil Palepu, Justin Chen, David Steiner, Jacqueline Shreibati, Wei-Hung Weng, Yilin Zhao, Xingjian Hu, Nicholas Zahn, Sadhya Garg, Julia Kirby, Yuxiang Gan, Jiaoli Li, Divy Thakkar, Shekoofeh Azizi, David Racz, Juraj Gottweis, Vivek Natarajan, Chenglin Wu, Tal Danino, Keran Rong, Haozhe Wang, Benoit Schillings, Yong Cheng, Quoc V. Le, Tao Tu
The paper introduces a SAREF-compliant ontology designed to represent distributed AI workflows across edge, fog, and cloud environments. It extends the SAREF4SYST ontology with concepts for AI pipelines, executable jobs, resources, deployment constraints, and communication links, creating a unified semantic model for both AI workflows and heterogeneous infrastructures. Evaluation through smart‑grid energy service scenarios and competency questions demonstrates successful deployment, reasoning, and workload adaptation, achieving 90‑100% deployment success and sub‑80 ms orchestration times.
By Viorica Rozina Chifu, Tudor Cioara, Vasile Ofrim, Liana Toderean, Ionut Anghel, Laura Daniele, Cornelis Bouter
DuMateBench is a new benchmark for autonomous agents that uses real user sessions from a large production platform, preserving interaction history, configurations, and workspace state. It contains 200 tasks across 8 scenarios and 17 capability categories, many requiring coordination of multiple capabilities. The benchmark tests agents in Docker containers with real-world complexities—Insufficient, Unstable, and Noisy—and evaluates performance with a hybrid deterministic and LLM-as-Judge protocol, revealing significant gaps in task completion across various agent frameworks and LLMs.
By Zechun Niu, Yukun Zhao, Jiaxin Zhang, Xu Shen, Jinhua Si, Han Tian, Can Xu, Yunfan Song, Jiaxin Mao, Yansong Gao, Yuchen Li, Jianmin Wu, Lingyong Yan, Shuaiqiang Wang, Dawei Yin
PILOT introduces a live self‑improvement framework for long‑horizon agents, enabling a supervisor to redirect or abort an active worker during execution and to distil learned procedures into reusable skills. Unlike existing single‑agent or subagent architectures, PILOT’s supervisor‑worker harness supports real‑time steering and self‑evolution. Experiments on two frozen backbones across three benchmarks show PILOT outperforming other harnesses, achieving significant gains in performance metrics and efficiency.
By Yang Xiao, Yusong Sun, Haoyi Wu, Wenyang Hui, Wen Da, Zhaokai Luo, Mu Chuan, Yao Hu, Wenjie Li, Chengyue Jiang
The paper examines whether large language model (LLM)-powered systems, such as OpenAI’s Swarm (OAS) framework, embody the core principles of classical swarm intelligence—decentralization, simplicity, emergence, and scalability. By implementing and comparing LLM-based and classical versions of Boids and Ant Colony Optimization, the authors find that while LLM swarms can mimic swarm-like dynamics, they suffer from significant computational overhead, with the LLM-based Boids simulation taking about 300 times longer than its classical counterpart.
By Muhammad Atta Ur Rahman, Melanie Schranz, Samira Hayat