概念
distillation on-policy training
创建2026-04-15
更新2026-04-15
阅读量级1 分钟
概念导读

Multi-domain on-policy distillation extends the OPD framework to handle multiple task domains simultaneously, training a single student model that can perform across diverse domains by distilling from domain-specific or generalist teachers. This concept is related to cascade-rl, which uses cascaded reinforcement learning for multi-domain post-training.

  1. Related
  2. References

Multi-Domain On-Policy Distillation

Multi-domain on-policy distillation extends the OPD framework to handle multiple task domains simultaneously, training a single student model that can perform across diverse domains by distilling from domain-specific or generalist teachers. This concept is related to cascade-rl, which uses cascaded reinforcement learning for multi-domain post-training.

References