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
- Vibe Coding a Data Pipeline Is Easy; Running It at Billions of Rows Is Not — Zachly · 2026-10-16
- How Claude Code Review works at GA: admin setup, auto reviews on every new PR — ClaudeDevs · 2026-10-16
- Anthropic offers eligible Claude Enterprise orgs at least $10,000 in Code Review credits — ClaudeDevs · 2026-10-16
- Anthropic makes Code Review generally available for Claude Team and Enterprise plans — ClaudeDevs · 2026-10-16
- Supabase CLI v2.119.0 Drops Docker, Gives Each Coding Agent Its Own Local Project — dshukertjr · 2026-10-16
- Musk amplifies user claim that Grok's computer use is fast enough to skip a Codex subscription — elonmusk · 2026-10-16