The authors study reinforcement learning for repository-level issue resolution, where an agent must interpret an issue, edit code, and validate patches over long trajectories.
Muon optimizer boosts training success for long-horizon repository agents
Combining Muon with single-rollout async updates, importance correction, and auxiliary penalties outperforms SPO baselines across two model families.
Research Lab
Kai Ruan · Jinghao Lin · Tao Chen · Ziqi Zhou · Qianshan Wei · Zihe Huang
Renmin University of China · Duke University · Institute of Automation, Chinese Academy of Sciences · Institute of Computing Technology, Chinese Academy of Sciences
Research Digest··2 min read
The authors evaluate a reinforcement-learning recipe for repository-level issue resolution that pairs the Muon optimizer with single-rollout asynchronous policy updates, token-averaged loss, truncated importance sampling, and auxiliary penalties.
Why this paper
From Institute of Automation, Chinese Academy of Sciences and 3 others · Part of RL for Tool Agents, now 23 papers
In one line
Muon optimizer combined with token-averaged loss and auxiliary penalties improves training success in long-horizon RL for repository-level issue resolution.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ·No stated limitations found
- ·No benchmark numbers found
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§