robert mccarthy

Rea­son­ing Models Strug­gle to Con­trol Their Chains of Thought

Open-weight train­ing prac­tices and im­pli­ca­tions for CoT monitorability

Early Signs of Stegano­graphic Ca­pa­bil­ities in Fron­tier LLMs

Can LLMs learn Stegano­graphic Rea­son­ing via RL?

[Paper] Hid­den in Plain Text: Emer­gence and Miti­ga­tion of Stegano­graphic Col­lu­sion in LLMs