作者位次 #11/11ACL 2026Findings· Findings of ACL
B alance SFT : Improving LLM Function Calling with Balanced Training Signals and Data Hardness
Bingguang Hao · Zengzhuang Xu · Maolin WANG · Yuntao Wen · Yicheng Chen · Cunyin Peng · Long Chen · an Jiaotong University) · Jinjie Gu · Chenyi Zhuang · Ji Zhang
摘要
While Supervised Fine-Tuning (SFT) is the prevailing method for equipping Large Language Models (LLMs) with function calling capabilities, its effectiveness is often compromised by two critical challenges: 1) Imbalanced Training Signals , where lengthy Chain-of-Thought (CoT) reasoning tokens domi...