作者位次 #3/10ACL 2026Poster· Natural Language Processing
B ase C al: Unsupervised Confidence Calibration via Base Model Signals
Hexiang Tan · Wanli Yang · Junwei Zhang · Xin Chen · Rui Tang · Du Su · Jingang Wang · Yuanzhuo Wang · Fei Sun · Xueqi Cheng
摘要
Reliable confidence is essential for trusting the outputs of LLMs, yet widely deployed post-trained LLMs (PoLLMs) typically compromise this trust with severe overconfidence. In contrast, we observe that their corresponding base LLMs often remain well-calibrated. This naturally motivates us to cal...