Beyond On-Policy vs. Off-Policy: At What Level Should LLM Agent Distillation Follow the Student?↗ (원문, 새 창)

본문 미리보기

arXiv:2610.31918v1 Announce Type: new Abstract: Distilling interactive LLM agents requires choosing between teacher-generated demonstrations and student-generated experience, corresponding broadly to off-policy and on-policy distillation. While both have proven effective, we argue that this distinction is inherently hierarchical: the policy that induces an interaction history need not be the same as the policy that generates the token prefix of the current action. We introduce a hierarchical ag