Stewart Slocum(@stewpervised):Could existing alignment audits have caught the behaviors in the OpenAI-HuggingFace incident before they happened? As a first step, we reproduced the incident with public models. 🧵

Stewart Slocum

@stewpervised

prev AI alignment @xai, phd @MIT

加入 September 2019

245 正在关注 1.4K 粉丝

Stewart Slocum@stewpervised

2026.09.12 13:01

Could existing alignment audits have caught the behaviors in the OpenAI-HuggingFace incident before they happened? As a first step, we reproduced the incident with public models. 🧵

显示更多

131

转发到社区

热门用户

0 粉丝

568.7K 粉丝

1.9M 粉丝

4.2M 粉丝

181.4K 粉丝

34K 粉丝

256.7K 粉丝

100.8K 粉丝

26.4M 粉丝

547.4K 粉丝

734.2K 粉丝

348.8K 粉丝

129.7K 粉丝

0 粉丝

23.7K 粉丝