English

Exploring Reasoning Reward Model for Agents

Artificial Intelligence 2026-04-29 v2 Computation and Language

Abstract

Agentic Reinforcement Learning (Agentic RL) has achieved notable success in enabling agents to perform complex reasoning and tool use. However, most methods still relies on sparse outcome-based reward for training. Such feedback fails to differentiate intermediate reasoning quality, leading to suboptimal training results. In this paper, we introduce Agent Reasoning Reward Model (Agent-RRM), a multi-faceted reward model that produces structured feedback for agentic trajectories, including (1) an explicit reasoning trace , (2) a focused critique that provides refinement guidance by highlighting reasoning flaws, and (3) an overall score that evaluates process performance. Leveraging these signals, we systematically investigate three integration strategies: Reagent-C (text-augmented refinement), Reagent-R (reward-augmented guidance), and Reagent-U (unified feedback integration). Extensive evaluations across 12 diverse benchmarks demonstrate that Reagent-U yields substantial performance leaps, achieving 43.7% on GAIA and 46.2% on WebWalkerQA, validating the effectiveness of our reasoning reward model and training schemes. Code, models, and datasets are all released to facilitate future research.

Keywords

Cite

@article{arxiv.2601.22154,
  title  = {Exploring Reasoning Reward Model for Agents},
  author = {Kaixuan Fan and Kaituo Feng and Manyuan Zhang and Tianshuo Peng and Zhixun Li and Yilei Jiang and Shuang Chen and Peng Pei and Xunliang Cai and Xiangyu Yue},
  journal= {arXiv preprint arXiv:2601.22154},
  year   = {2026}
}

Comments

ACL 2026 Findings, Project page: https://github.com/kxfan2002/Reagent

R2 v1 2026-07-01T09:26:27.524Z