RSS

Yavuz Bakman

Karma: 30

Thinking about AI Alignment and Reliability.

Enjoying Soulsborne games.

LLM Misal­ign­ment Can be One Gra­di­ent Step Away, and Black­box Eval­u­a­tion Can­not De­tect It.

Yavuz Bakman15 Mar 2026 0:19 UTC
31 points
4 comments3 min readLW link