AI RESEARCH
Large Language Models Are Overconfident in Their Own Responses
arXiv CS.CL
•
ArXi:2606.03437v1 Announce Type: new Prior work has shown that instruction-tuned large language models (LLMs) are less well calibrated than their base pre-trained counterparts. However, little is known about the frequently used chat template's effect on the calibration of conversational LLMs. In this work, we investigate the mechanisms driving this miscalibration by decoupling the effects of the post-