The paper presents a multi‑agent, persona‑driven NL2SQL framework that addresses common LLM shortcomings such as schema hallucination, syntax errors, and intent mismatch. It incorporates domain relevance detection, schema‑grounded knowledge construction, KPI‑aware SQL synthesis, syntax‑level self‑repair via sqlite feedback, and semantic consistency checking through SQL‑to‑NL reverse engineering. Evaluated on 60 queries, the system achieves 90% correctness using Mistral across seven coordinated LLM calls, demonstrating a significant improvement in reliability for high‑stakes environments.
By Anjali Vemula, Akshat Dasula
An agentic AI system for automated troubleshooting in optical communication networks is demonstrated that integrates data references into the agent architecture to prevent data hallucinations. This integration improves the interpretability and trustworthiness of root‑cause identification and reasoning. The approach addresses a key challenge in deploying AI for network maintenance by reducing erroneous outputs.
By Jeremias D\"otterl, Martin Kuipers
MobileInject is a benchmark that evaluates the security of multimodal large language model (MLLM)-based mobile agents against multimodal injection attacks. It offers a high‑fidelity interactive framework with 11 scenarios, 9 risk categories, and 108 executable risk instances, and introduces Cross‑Modal Consistency Injection along with two advanced attack methods—Multi‑Role Collaborative Deception and Reasoning‑Camouflage Injection—to intensify trust manipulation and reasoning hijacking. Experiments show that mobile agents are broadly vulnerable, especially to early‑stage reasoning hijacking, establishing MobileInject as a systematic tool for studying and improving mobile agent security.
By Songping Wang, Peiyang Li, Yi Huang, Fengting Li, Xuhao Ran, Yueming Lyu, Chao Feng, Caifeng Shan
The paper introduces JointCoreBench, a benchmark for evaluating memory selection in large language model agents, and JointCore, a cost‑ordered group‑deletion baseline that identifies minimal sufficient memory subsets. JointCore achieves perfect recovery of minimum‑cost sufficient sets across 6,912 conditions, cuts subset evaluations by 62.9%, and saves 87.8% of active memory usage. Experiments across multiple model families and tasks demonstrate that JointCore preserves full‑memory behavior while reducing model queries and cost‑aware recovery rates compared to singleton deletion.
By BHVSP Subrahmanyam, Neeraj Choudhary
GLOBE is an agentic framework designed to improve e‑commerce product listings for generative search by storing optimization knowledge in an evolving handbook of explicit, field‑scoped behaviors. In each episode, an Editor applies relevant behaviors, receives rank‑based feedback, and a Reflector converts this feedback into behavior updates. A Moderator removes redundant proposals, a Curator commits accepted changes, and harmful behaviors are deactivated, turning episodic feedback into a persistent, field‑scoped optimization memory that outperforms monolithic prompt optimization and remains effective on unseen product categories.
By Hyunseo Kim, Suil Choe, Sukkyoung So, Daeyeon Yi
A survey paper titled "A Survey of Explainable Reinforcement Learning: Targets, Methods and Needs" reviews over 250 studies in the sub‑domain of eXplainable AI focused on reinforcement learning (XRL). The authors introduce a taxonomy based on the questions "What" (the target of explanation) and "How" (the method of explanation) and use it to organize the state‑of‑the‑art literature. They also highlight related domains that warrant further attention and outline key needs for advancing the field of XRL.
By L\'eo Sauli\`eres
EgoP$^{2}$C is a new framework that internalizes a plan‑perceive‑compose workflow within a single vision‑language model for spatial reasoning in egocentric video. It trains specialized Planner and Perception teachers via Dual‑LoRA adapters, then fine‑tunes a mixture‑of‑experts student under their supervision, using a segment‑level multi‑teacher on‑policy distillation to guide evidence‑seeking, stopping, and answer composition. The resulting student models, built on Qwen3‑VL‑8B and SenseNova‑SI‑1.3‑8B, outperform their base models by 9.2 and 7.5 percentage points on three benchmarks, while cutting inference FLOPs by about 66% and speeding up inference 1.29–1.49×.
By Chaoyu Li, Arka Sadhu, Yuliang Li, Hyo Jin Kim, Yuning Chai, Pooyan Fazli
Q-Reflector is a standalone GUI action reflector built on Qwen3‑VL‑8B that learns to generate pre‑ and post‑conditions for each executed action and then evaluates execution success and task validity. It is trained with supervised fine‑tuning, teacher‑student distillation, and reward‑based refinement, and achieves 82.16% accuracy on 1,082 AndroidWorld transitions, outperforming baseline models. In MobileWorld tasks, it improves task success rates by 17.09 and 14.53 percentage points over no external reflection, and surpasses GPT‑5.2 and Claude Haiku 4.5 by 3.42 and 5.98 percentage points respectively.
By Linqiang Guo (Peter), Wei Liu (Peter), Li Gu (Peter), Yang Wang (Peter), Tse-Hsun (Peter), Chen
The paper introduces CrEST, a hierarchical credit assignment framework for multi-turn, multi-step large language model agents. CrEST anchors each token’s advantage sign to verifier-defined rewards while using a self-teacher to provide dense magnitude signals, addressing both inter-turn dilution and intra-turn token contribution issues. Experiments on BFCL V3 and WildToolBench show CrEST consistently outperforms reinforcement learning and distillation baselines, especially on long trajectories and strict session-level metrics.
By Zechuan Wang, Siyuan Lu, Hongxuan Zhang, Linjian Mo, Chenyi Zhuang, Leilei Gan
EvalResearchBench (ERB) is a benchmark that tests whether AI agents can autonomously design evaluation tasks for recursive self‑improvement. In ERB, a researcher agent selects or synthesizes tasks, builds graders, and refines them through pilot tests before producing a final evaluator that runs within fixed time and API budgets. Experiments with nine researchers and thirteen candidate models show that the best automated evaluators achieve about 75% concordance with target benchmarks, below the 91% ceiling set by target disagreements, and that human‑designed task samples still outperform many automated approaches.
By Yaolun Zhang, Tianyi Xu, Yujie Zhao, Jishen Zhao, Qingyun Wu, Huazheng Wang
The paper investigates how the representation of tool interfaces—referred to as a tool schema—affects the performance of large language model agents. By systematically transforming a native tool schema with nine operators, the authors demonstrate that agents exhibit significant variability in success rates across different but functionally equivalent schema variants, a phenomenon they call schema bias. Experiments on eleven LLMs, including closed models, reveal that success can range from complete failure to 97% solely due to schema differences, and that training only mitigates bias when the specific variant is included in the training data.
By Yinhong Liu, Zhili Tan, Zilin Wang, Zhijiang Guo
The study examined how AI agents affect teamwork, productivity, and performance using a large-scale experiment on the Pairit platform. Human-AI teams produced 50% more ads per worker and higher text quality, while human-human teams produced higher image quality. The results showed that task-oriented communication and delegation in human-AI teams improved productivity and quality across all dimensions, whereas interpersonal communication benefited human-human teams.
By Harang Ju, Sinan Aral
The study examines whether large language models (LLMs) can accurately emulate human goal selection in a self‑directed learning task. Five models—GPT‑5, Gemini 2.5 Pro, Claude Sonnet 4.5, Qwen3 32B, and Centaur—were tested and found to diverge significantly from human behavior, often exploiting a single solution or performing poorly, with little variation across instances. Techniques such as chain‑of‑thought reasoning and persona steering offered only limited improvements, and the results were consistent across different experimental settings.
By Gaia Molinaro, Dave August, Danielle Perszyk, Anne G. E. Collins
The paper introduces AOI (Autonomous Operations Intelligence), a multi‑agent framework that treats automated site reliability tasks as a structured trajectory learning problem under security constraints. It combines a trainable diagnostic system using Group Relative Policy Optimization, a read‑write separated execution architecture, and a Failure Trajectory Closed‑Loop Evolver that turns unsuccessful runs into corrective supervision. On the AIOpsLab benchmark, AOI achieves significant performance gains, including a 66.3% best@5 success rate and improved average metrics on unseen fault types.
By Pei Yang, Wanyi Chen, Asuka Yuxi Zheng, Xueqian Li, Xiang Li, Haoqin Tu, Jie Xiao, Yifan Pang, Dongdong Zhang, Fuqiang Li, Alfred Long, Lynn Ai, Eric Yang, Bill Shi, Yifan Chen, Shuhan Sun
The paper introduces a process‑aware evaluation framework for cooperative multi‑agent reinforcement learning that supplements traditional outcome metrics with diagnostics of interference, coordinated task completion, participation, and coordination efficiency. Using the STAT testbed and two established benchmarks, the authors demonstrate that policies with similar returns can differ markedly in coordination behavior, and that reward shaping can alter coordination without affecting overall task return. The framework reveals distinct process signatures across spatial, task, and agent scaling and remains informative across different coordination mechanisms.
By Maria Ana Cardei, Matthew Landers, Afsaneh Doryab
Lean4AgenticRL is a new framework that introduces multi‑level process rewards grounded in the Lean4 formalization schema to improve agentic reinforcement learning for large language models. It constructs a problem‑specific workflow system that balances capability and controllability, enabling verification of agent behavior and providing rewards at low‑, mid‑, and high‑levels. Experiments show that models trained with Lean4AgenticRL, such as LAR‑27B, outperform comparable models, achieving 77.3% on TerminalBench‑2.1 and surpassing the base model by an average of 9.89%.
By Ruida Wang, Gerald Friedland, Tong Zhang, Chris Kong
MetaCurate is a training‑free framework that automatically evolves curation filters for agentic coding data by using pairwise large language model judgments to refine its rules, rubric, and aggregation policy. Starting from no initial filter, it discovers a strategy that scores trajectories independently, requiring pairwise comparisons only during the search phase. Experiments on an uncurated trajectory pool and the Nemotron v3 SWE dataset show that MetaCurate achieves the highest mean SWE‑Verified score compared to random selection, hand‑authored filtering, and a training‑loop baseline, while incurring lower search costs.
By Donghoon Kim, Seunghyuk Oh, Kevin Galim, Minseo Kim, Minjae Lee, Younghyun Kim, Wooseok Seo, Hyung Il Koo, Wonjun Kang
FLARE is a physics‑grounded benchmark for evaluating language‑model agents on feeder restoration tasks after distribution‑grid faults. It includes 800 AC‑feasible cases across five observation settings, allowing agents to gather evidence, evaluate switches, and perform restoration actions under fixed operational limits. In tests of five instruction‑tuned models over 78,750 episodes, mean normalized regret ranged from 0.082 to 0.148, with a high P95 regret indicating a significant failure tail, and 96.4% of guard‑rejected actions would have caused physical violations.
By BHVSP Subrahmanyam, Neeraj Choudhary
The paper introduces SPLIT, a method that optimally allocates test‑time compute among sampling, model selection, and verification for language models and agents. Using an exact probability formula that accounts for task difficulty, shared blind spots, and verifier correlations, SPLIT is fitted via empirical Bayes on a small labeled pilot and then used to plan budgets for unseen tasks. Across seven diverse benchmarks, 26 models, and six verifier types, SPLIT achieves near‑optimal performance, outperforming rule‑based baselines by up to 2.5 percentage points and providing insights into when additional LLM judges reduce precision.
By Srinivasan Iyengar
CocoRNA is a cooperative multi-agent reinforcement learning framework that tackles RNA inverse design by breaking the problem into smaller sub‑problems solved collaboratively by multiple agents. A centralized critic provides global structural guidance during training, allowing the agents to jointly optimize their design strategies. Experiments on the Rfam dataset show that CocoRNA outperforms state‑of‑the‑art methods in both success rate and design speed, and additional tests on other biological sequence design tasks demonstrate its broad effectiveness.
By Tianmeng Hu, Biao Luo, Ke Li