LLMs have a "truth vector" - an emergent linear structure that represents factual truth values

LLMs contain a specific "truth direction" denoting factual truth values

False information and true information cluster separately in LLMs. From the paper.

Artificial intelligence systems like large language models have shown impressive capabilities, such as engaging in conversation, answering questions, and generating coherent text. However, they are also prone to making clearly false statements or hallucinating incorrect in…