Hugging Face Hub now hosts 88 runnable RL environment datasets for agent training
mervenoyann · x · 2026-09-30
RL Environments Come to HF Hub
Hugging Face engineer mervenoyann announced that RL environments can now ship as datasets on the Hub, with a new rl-environment filter listing 88 environments so far.
- Environments are ready to run with Harbor, Verifiers, OpenEnv, and NeMo Gym
- Cataloged environments include nvidia's Nemotron series (instruction following, agent calendar scheduling, ReasoningGym-v1), FineEnvs' SmolDataEnvs, MiMo-V2.6 RL-harbor-cyber, and more
- Datasets support /csv/parquet/arrow formats plus agent traces, filterable alongside benchmarks
This gives agent RL training a centralized distribution channel — an "environments as datasets" ecosystem akin to model weights.
More from coding & agent
- LangChain launches LangSmith Engine v2 with proactive red-teaming for agents — LangChain · 2026-10-01
- HQ launches as shared AI context layer for humans and agents, with ~1,000 businesses onboard — jacob_posel · 2026-10-01
- Agents Are Taking Over Data: A Conversation with dbt Founder Tristan Handy — juansequeda · 2026-10-01
- NVIDIA Unveils a New Way to Build Edge AI with Agentic Development on Jetson — NVIDIAAI · 2026-09-30
- Hard checks on tool calls beat prompt rules: 4 agent guardrail patterns from 8,176-call replay — Individual-Shower973 · 2026-09-30
- "AI Software Factory" Concept Gains Traction: Faster Coding Just Moves the Bottleneck — Pavan_Belagatti · 2026-09-30