Taywon Min

Working on AI safety & AI alignment at KAIST AI.

[Shi Feng @ihsgnef]

New post: Sycophancy Towards Researchers Drives Performative Misalignment

We found no clear evidence that scheming is more valid than sycophancy to explain alignment faking. 🧵

[Juheon Choi @_JuheonChoi_]

Often distracted by shorts, videos, and digital noise? Introducing INA, an AI agent for intentional living. State your intention — the agent guides you back when you drift off-task. Paper:
https://t.co/qDR05lRLgw
Website & code (Mac app): https://t.co/3QDiyKcBsh
Thread below 1/N

[Taywon Min @MinTaywon]

@emollick One possible (imo, understudied) approach to mitigate sycophancy would be to develop better, systematic methods identifying sycophantic labels in preference datasets. We invesitgated this direction in our work accepted at ACL 2025.
check it out! https://t.co/0FjLEAporw

[Taywon Min @MinTaywon]

@ErnestRyu 정말 좋은 인사이트인 것 같습니다. 서울대에 좋은 교수님들이 더 많이 오셔서 학생들에게도 다양한 연구분야를 접할 수 있는 기회들이 더욱 늘어나면 좋겠습니다ㅎㅎ

[Taywon Min @MinTaywon]

@ryangreenblatt This post is amazing! Really like the open way that you share these nice ideas. Especially interested in the reward hacking proposals