概念
rlhf grpo training
创建2026-04-15
更新2026-04-15
阅读量级1 分钟
概念导读

Free Process Rewards refers to the idea of obtaining process-level reinforcement learning reward signals without requiring manual annotation. This concept is explored in generalized-on-policy-distillation as a way to provide richer training signals for on-policy distillation. By leveraging existing model outputs or automated verification, free process reward...

  1. Related
  2. References

Free Process Rewards

Free Process Rewards refers to the idea of obtaining process-level reinforcement learning reward signals without requiring manual annotation. This concept is explored in generalized-on-policy-distillation as a way to provide richer training signals for on-policy distillation. By leveraging existing model outputs or automated verification, free process rewards aim to reduce the cost of RL training while maintaining process-level granularity.

References