Hugging Face TRL makes score centering + masked importance sampling default in async GRPO
_lewtun · x · 2026-10-27
Hugging Face's TRL team announced that score centering + masked importance sampling is now the default in the async GRPO Trainer, showing remarkably stable training across various staleness and sampler quantization settings.
A practical win for teams running LLM RL post-training: staleness and quantized-sampling instability in asynchronous GRPO now has an out-of-the-box mitigation, no custom tuning required.
More from coding & agent
- Agent Field Kit launches $29 security runbooks for production AI agents — alvelda · 2026-10-27
- Dev mixes Claude specs with abliterated GLM to build SS7 threat detector — l4rz · 2026-10-27
- Cloudflare Durable Objects validated as top architecture for AI agent WebSocket servers — tristanbob · 2026-10-27
- GitHub Next's Chopin prototype explores real-time multiplayer agentic planning — max__drake · 2026-10-27
- Veynod: An Open-Source Shared Workspace Where AI Agents Collaborate Like a Team — TKOWNS · 2026-10-27
- TMLR paper: enforced thinking doesn't reliably make LLM agents disclose key info to users — SharonYixuanLi · 2026-10-27