Adithya S K(@adithya_s_k):1/ Excited to release The ultimate guide to multi-harness RL The same model behaves differently in every agent harness. So we built an open way to train any model with RL on any task set, inside the harnesses people actually use, like Claude Code, Codex, and OpenCode, without changing a single line of harness or training code. Trained across four harnesses, LFM2.5-2.6B went from 42% to 54% with 31% fewer tool calls. 🧵

Adithya S K

@adithya_s_k

Scaling RL Envs @huggingface 🤗 • Founded @cognitivelab_ai Prev : Research @MSFTResearch • ML @apple • 22

加入 June 2020

2.5K 正在关注 16.6K 粉丝

Adithya S K@adithya_s_k

2026.10.01 15:43

1/ Excited to release The ultimate guide to multi-harness RL The same model behaves differently in every agent harness. So we built an open way to train any model with RL on any task set, inside the harnesses people actually use, like Claude Code, Codex, and OpenCode, without changing a single line of harness or training code. Trained across four harnesses, LFM2.5-2.6B went from 42% to 54% with 31% fewer tool calls. 🧵

显示更多