Agent0 is a fully autonomous framework that enables large language model agents to evolve without external data by using a multi‑step co‑evolution process. It pits a curriculum agent against an executor agent, both derived from the same base LLM, where the curriculum agent creates increasingly challenging tasks and the executor learns to solve them. By integrating external tools into the executor’s workflow, the system creates a self‑reinforcing cycle that continuously generates high‑quality curricula, leading to significant gains in reasoning performance—an 18% improvement on mathematical reasoning and 24% on general reasoning for the Qwen3‑8B‑Base model.
By Peng Xia, Kaide Zeng, Jiaqi Liu, Can Qin, Fang Wu, Yiyang Zhou, Caiming Xiong, Huaxiu Yao
arXiv:2509. 26306v5 Announce Type: replace Abstract: Existing multi-agent learning approaches have developed interactive training environments to explicitly promote collaboration among multiple Large Language Models (LLMs), thereby constructing stronger multi-agent systems (MAS).
By Hehai Lin, Shilei Cao, Sudong Wang, Haotian Wu, Minzhi Li, Linyi Yang, Juepeng Zheng, Chengwei Qin
arXiv:2601. 03555v3 Announce Type: replace Abstract: Training reliable tool-augmented agents remains a significant challenge, largely due to the difficulty of credit assignment in multi-step reasoning.
By Yuxuan Jiang, Francis Ferraro
SynCo is a data synthesis co‑training framework that enables self‑evolving large language models (LLMs) to improve autonomously. It employs two agents—a Synthesizer that generates training tasks based on the Reasoner’s evolving capabilities, and a Reasoner that learns from these tasks—using multi‑agent reinforcement learning to jointly optimize task quality and model performance. Experiments on eight mathematical reasoning benchmarks show that SynCo outperforms existing synthetic‑data methods, especially on previously unsolved problems.
By Wei Yang, Shawn Li, Yuehan Qin, Yawei Wang, Mingxi Wang, Shixuan Li, Tiankai Yang, Jiate Li, Jesse Thomason, Xuezhe Ma, Yue Zhao
arXiv:2510. 05592v2 Announce Type: replace Abstract: Outcome-driven reinforcement learning has advanced reasoning in large language models (LLMs), but prevailing tool-augmented approaches train a single, monolithic policy that interleaves thoughts and tool calls under full context; this scales poorly with long horizons and diverse tools and generalizes weakly to new scenarios.
By Zhuofeng Li, Haoxiang Zhang, Seungju Han, Sheng Liu, Jianwen Xie, Yu Zhang, Yejin Choi, James Zou, Pan Lu
arXiv:2610.19868v1 Announce Type: cross
Abstract: Training LLM agents on long-horizon interactive tasks with reinforcement learning typically relies on human-annotated tasks and reliable verification...
By Tianyu Liu, Davide Belli, Bence Major, Arash Behboodi, Tribhuvanesh Orekondy
arXiv:2512. 03438v3 Announce Type: replace Abstract: Agentic reasoning models trained with multimodal reinforcement learning (MMRL) have become increasingly capable, yet they are almost universally optimized using sparse, outcome-based rewards computed based on the final answers.
By Reuben Tan, Baolin Peng, Zhengyuan Yang, Hao Cheng, Oier Mees, Theodore Zhao, Andrea Tupini, Isar Meijer, Qianhui Wu, Yuncong Yang, Lars Liden, Yu Gu, Sheng Zhang, Xiaodong Liu, Lijuan Wang, Marc Pollefeys, Yong Jae Lee, Jianfeng Gao
The paper introduces Verifiable Process Rewards (VPR), a framework that uses symbolic or algorithmic task-solving oracles as action-level verifiers to provide dense rewards for intermediate actions in agentic reasoning tasks. VPR is applied to Sokoban, Sudoku, and Minesweeper, and improves policy performance over outcome-reward baselines. The authors also demonstrate VPR’s effectiveness in a tool‑use benchmark, highlighting its reliance on reliable task oracles and the difficulty of extending it to less structured environments.
By Huining Yuan, Zelai Xu, Huaijie Wang, Xiangmin Yi, Jiaxuan Gao, Xiao-Ping Zhang, Yu Wang, Chao Yu, Yi Wu
The paper introduces MSARL, a Multi‑Small‑Agent Reinforcement Learning framework that separates reasoning from tool use. A single Reasoning Agent plans and decomposes tasks, while multiple specialized Tool Agents handle specific external tools, each trained with imitation learning and reinforcement learning using role‑specific rewards. Experiments on mathematical problem solving with code execution show that MSARL improves reasoning stability and final‑answer accuracy compared to single‑agent baselines, and the architecture generalizes to other tool‑use tasks.
By Dayu Wang, Yutong Liu, Jiaye Yang, Weikang Li, Jiahui Liang, Yang Li
arXiv:2512. 09706v2 Announce Type: replace Abstract: The paradigm of agentic AI is shifting from engineered complex workflows to post-training native models.
By Kaichen He, Zihao Wang, Muyao Li, Anji Liu, Yitao Liang
The paper introduces a text-based, multi-agent version of the board game Clue to test multi-step deductive reasoning in large language models (LLMs). Six LLM-based agents (GPT‑4o‑mini and Gemini‑2.5‑Flash) play turn‑based games, and a tool‑augmented approach uses a structured possibility matrix to convert implicit game state into explicit remaining possibilities, thereby offloading memory and deductive constraints from the agents. The study compares this tool‑augmented method against a baseline to assess its impact on reasoning quality and task success in a strategic reasoning environment.
By Rebecca Ansell, Autumn Toney-Wails
OmniBrain-0 is a lightweight, unified embodied agent that coordinates manipulation and navigation through a shared high‑level decision model. It integrates task understanding, planning, progress tracking, and feedback‑driven recovery, and is trained via embodied pretraining, domain‑specialized agent distillation, and multi‑teacher on‑policy distillation. The framework also introduces Rubric‑Guided Recursive Self‑Improvement, a data flywheel that refines rubrics and trajectories for continual domain‑expert updates, enabling efficient transfer and consolidation of embodied capabilities across multiple benchmarks and real‑world tasks.
By Xiaomi Embodied Intelligence Team, Yuechen Luo (University of Macau, Tsinghua University), Chong He (University of Macau, Tsinghua University), Fang Li (University of Macau, Tsinghua University), Shaoqing Xu (University of Macau, Tsinghua University), Jingkai Wang (University of Macau, Tsinghua University), Menglun Su (University of Macau, Tsinghua University), Jiaxin Liu (University of Macau, Tsinghua University), Shengyin Jiang (University of Macau, Tsinghua University), Zhiyi Lai (University of Macau, Tsinghua University), Long Chen (University of Macau, Tsinghua University), Longlong Wang (University of Macau, Tsinghua University), Liang Hu (University of Macau, Tsinghua University), Haochen Liu (University of Macau, Tsinghua University), Yuncheng Jiang (University of Macau, Tsinghua University), Hongwei Xie (University of Macau, Tsinghua University), Jieming Cui (University of Macau, Tsinghua University), Hanbing Li (University of Macau, Tsinghua University), Junjie Huang (University of Macau, Tsinghua University), Zhixiang Duan (University of Macau, Tsinghua University), Hexin Zhang (University of Macau, Tsinghua University), Weixiang Liang (University of Macau, Tsinghua University), Ying Gong (University of Macau, Tsinghua University), Yuhan Wang (University of Macau, Tsinghua University), Yong Pan (University of Macau, Tsinghua University), Pengna Li (University of Macau, Tsinghua University), Kangyi Wu (University of Macau, Tsinghua University), Ken Deng (University of Macau, Tsinghua University), Kun Ma (University of Macau, Tsinghua University), Fuxi Wen (University of Macau, Tsinghua University), Zhi-xin Yang (University of Macau, Tsinghua University), Hangjun Ye (University of Macau, Tsinghua University)