Iterative GRPO: Batch-Online Multi-Turn RL via Single-Turn RLHF
Practical LLM agents often operate over multi-turn conversations where success is determined only after the full interaction ends. Most multi-turn RL methods train via on-policy rollouts, but unlike in single-turn RLHF, the policy cannot produce a trajectory alone, since an external environment must respond after each...
Daniel Jiang, Ankur Samanta, Yukai Yang et al.
· 2 citations