作者位次 #2/6ACL 2026Poster· Natural Language Processing
Select Before Use: On the Importance of Reference Model Selection in Preference Alignment
Muyang Li · Runze Wu · an Jiaotong University) · Bo Han · Daoyi Dong · Tongliang Liu
摘要
The post-training stage of Large Language Models (LLMs) typically involves Supervised Fine-Tuning (SFT) followed by preference alignment to ensure LLM to generate safe, helpful, and instruction-aligned content. The SFT model critically serves as both the initialization and reference model for sub...