ClawGym II paper: Improving agents via mixed-harness training
omarsar0 · x · 2026-08-19
The paper 'ClawGym II' introduces a method to train agents using existing code harnesses like OpenClaw and Claude Code. By placing a serving proxy at the model boundary, the system captures tool calls and organizes them into prefix trees, allowing PPO and GRPO to optimize over the multi-turn structure. Results show Qwen3-30A3B gained significant Pass@1 improvements (+9.98 via OpenClaw, +14.81 via Claude Code). The study also demonstrates that mixed-harness training leads to policies that generalize across different execution systems rather than overfitting to a single one.
Related event: RUC's ClawGym II Trains Agents via Black-Box RL on Real Harnesses(3 posts)→
More from coding & agent
- React Grab boosts agent coding 2x by mapping UI to source — aidenybai · 2026-08-19
- Agent skill anti-slop flags 202 errors in a single repo test — soumitrashukla9 · 2026-08-19
- React Grab adds support for Three.js and React Three Fiber — aidenybai · 2026-08-19
- High-quality foundational libraries are the overlooked key to long-term engineering success — wavefnx · 2026-08-19
- Grok Bot Under the Hood: Managed VM Agent Control — noahsolomon · 2026-08-19
- Proposal: SKILL.md format needs #ifdefs for harness — mertdumenci · 2026-08-19