Training Agents IRL: GitHub and Hugging Face host Open Source AI Week event on multi-step agent RL
ben_burtenshaw · x · 2026-09-29
The Open Intelligence Foundation announced Training Agents IRL, an Open Source AI Week event co-hosted by GitHub and Hugging Face on October 22 in San Francisco.
The event goes beyond single-turn RL to explore how agents learn and act across multi-step environments, covering:
- Reward design
- Rollouts and environment-driven training
- Benchmarks and evaluation
- The gap between training gains and real-world behavior
Short talks will come from researchers and builders at Hugging Face, Microsoft, Mercor, Reflection, RadixArk, and Benchflow, followed by a panel on what works, bottlenecks, and what the open-source ecosystem needs next.
More from coding & agent
- PHP WASM kernel hits major milestone with PR #936 merge, author vows to keep backing open source — MickeySteamboat · 2026-09-29
- Google to replace Gemini Gems with Skills starting November 17 — mark_k · 2026-09-29
- Carla v0.1.0: a local llama.cpp loom TUI for growing AI characters — max_paperclips · 2026-09-29
- Running Jev at high frame rate with full-state snap inferences makes it a true System 1 — mathemagic1an · 2026-09-29
- Meeting agent memory design: retain facts instead of patching tags — Extension-Answer4821 · 2026-09-29
- MegaMemory: a persistent knowledge graph MCP server that lets coding agents remember across sessions — tom_doerr · 2026-09-29