作者位次 #4/8ACL 2026Findings· Findings of ACL
E^3 - TIR : Enhanced Experience Exploitation for Tool-Integrated Reasoning
Weiyang Guo · Zesheng Shi · Liye Zhao · Jiayuan Ma · Zeen Zhu · Junxian He · Min Zhang · Jing Li
摘要
While Large Language Models (LLMs) have demonstrated significant potential in Tool-Integrated Reasoning (TIR), existing training paradigms face significant limitations: Zero-RL suffers from inefficient exploration and mode degradation due to a lack of prior guidance, while SFT-then-RL is limited...