Human Folly about “abliteration”

index

• Refusal in LLMs is mediated by a single direction (lesswrong.com):

  Arditi et al. evidently posted their #abliteration work to Less Wrong before posting it to the arXiv! #AI #neural-networks

  on 02026-05-17
• (arxiv.org):

  Arditi et al.’s “Refusal in language models is mediated by a single direction” #paper. #AI #neural-networks #abliteration

  on 02026-05-17
• Abliteration (huggingface.co):

  explaining how #abliteration of #neural-networks works, based on Arditi et al.’s “Refusal in language models is mediated by a single direction” paper. #AI

  on 02026-05-17