[2605.30381] When LLMs Learn to Be Consistently Wrong: A Multi-Model Study of Linear Representations of Synthetic Deception
![[2605.30381] When LLMs Learn to Be Consistently Wrong: A Multi-Model Study of Linear Representations of Synthetic Deception](/static/img/curated/afbytes-us.png)
AI disclosure
Summary
Abstract page for arXiv paper 2605.30381: When LLMs Learn to Be Consistently Wrong: A Multi-Model Study of Linear Representations of Synthetic Deception