Hugging Face synthesizes RL envs from open trajectories: Repair, Build, Extend, QA
yb2698 · x · 2026-11-06
To increase difficulty and diversity beyond previous data-analysis-focused RL environments, Hugging Face used public agentic trajectories on the HF Hub as seed data, converting them into tasks only when they contain a verifiable unit of work. Four new categories:
- Repair: a check that failed, then passed after edits
- Build: a project finished from one request
- Extend: a later change whose checks pass
- QA: a fact computed from public data
Each task is validated by running its oracle in its own container image, kept only if the reference agent (LFM2.5-24B) solves it sometimes but not always — ensuring a learning signal for GRPO.
More from coding & agent
- One plain-English prompt turns your phone into a 200-particle physics simulator — tak3sh8 · 2026-11-06
- Zep Agent Skills turns verified agent traces into versioned, governed procedural memory — TheTuringPost · 2026-11-06
- Zep launches Agent Skills: compile verified agent runs into reusable skills — TheTuringPost · 2026-11-06
- Sentry's next MCP evolution: one CLI login powers both MCP and full CLI access — zeeg · 2026-11-06
- Box launches Box Content Retrieval with agentic platform and MCP support at BoxWorks — shashib · 2026-11-06
- Give your agent an interpreter, not tools: 19 min to 2 min on the same model — Itz_hany · 2026-11-06