Back to Home
MIT CSAIL··Papers & Tech

Teaching AI models to say “I’m not sure”

中文摘要

新训练方法提升AI置信度评估能力以减少幻觉现象

English Summary

A new training method improves AI confidence estimation, allowing models to express uncertainty and reducing hallucinations without sacrificing performance.

Original Excerpt

A new training method improves the reliability of AI confidence estimates without sacrificing performance, addressing a root cause of hallucination in reasoning models.