HuggingFace TRL Adds Native Support for MOPD

Hugging Face's TRL library now natively supports MOPD (Multi-Teacher On-Policy Distillation), a trending post-training technique where multiple RL expert models are distilled into one; several frontier open-source models already use this route.

2026-11-03 ~ 2026-11-04 · 2 related posts

• MOPD becomes the post-training recipe of the year: TRL now ships multi-teacher on-policy distillation — SergioPaniego · 2026-11-03
• TRL adds native support for MOPD, the distillation trick frontier open models use to merge RL experts — pcuenq · 2026-11-04