SkillRL (NeurIPS 2026): 7B model beats GPT-4o by 41% via recursive skill evolution
cihangxie · x · 2026-09-25
SkillRL (arXiv:2602.08234), accepted to NeurIPS 2026, tackles LLM agents' failure to accumulate skills from experience. It proposes recursive skill-augmented RL: experience-based distillation builds a hierarchical skill library (SkillBank), an adaptive retrieval strategy covers general and task-specific heuristics, and a recursive evolution mechanism lets the skill library co-evolve with the policy during RL — cutting token footprint while boosting reasoning utility.
On ALFWorld, WebShop, and seven search-augmented tasks it sets SOTA: a 7B model outperforms GPT-4o by 41%, with 20% fewer training tokens and 33% faster convergence. Paper, data, and models are open-sourced on arXiv, HuggingFace, and GitHub.
More from coding & agent
- Docker launches Cloud Sandboxes: microVM isolation for always-on agents, $250 credit — juntao · 2026-09-25
- Vercel AI SDK to add OpenAI functions support with three UX flows — maxleiter · 2026-09-25
- Developers now prompt AI even to change a CSS color code — saheedniyi_02 · 2026-09-25
- GPT-6 "Astra" agent clears three chapters of Left 4 Dead 2 with 0/5 escapes — imjustnewatai · 2026-09-25
- Perplexity launches Rust-based Fast Search: 95% of results in 230ms, Shopify adopts it — AravSrinivas · 2026-09-25
- Gergely Orosz: knowing how to code makes AI 95-99% faster than typing — ducha_aiki · 2026-09-25