Skip to content
arXiv · Computation and Language· Junnan Liu, Linhao Luo, Zhijun Chen, Qianren Mao, Thuy-Trang Vu, Gholamreza Haffari·· 4 hr agoAI score30

STI-OPD:面向智能体在线策略蒸馏的随机教师干预框架

Stochastic Teacher Intervention for Agentic On-Policy Distillation

AI brief

研究者提出 STI-OPD,一种面向多轮智能体在线策略蒸馏(OPD)的随机教师干预框架,用教师-学生策略差异引导替换学生动作,以获取更可靠的 token 级监督。

Source: arXiv · Computation and Language · arxiv.org