作者位次 #3/3ACL 2026Poster· Natural Language Processing
Can Compact Language Models Search Like Agents? Distillation-Guided Policy Optimization for Preserving Agentic RAG Capabilities
Rikuto Kotoge · Mai Nishimura · Jiaxin Ma
摘要
Reinforcement Learning has emerged as a dominant post-training approach to elicit agentic RAG behaviors such as search and planning from language models. Despite its success with larger models, applying RL to compact models (e.g., 0.5-1B parameters) presents unique challenges. The compact models...