Hugging Face Papers ↗ · 2026-11-14
智能体策略内蒸馏隐式重塑了 LLM 的 ReAct 模式
Media reports Hugging Face Papers Source language: Chinese
论文摘要(作者报告,未经本站复测):策略内蒸馏(OPD)将专业能力迁移到语言模型智能体,但其对交互的跨领域影响仍不清楚。
Full text
On-policy distillation can improve or degrade language-model agent interaction across domains, and selectively retaining teacher-preferred positions mitigates negative transfer at a cost to source-domain learning.