C cdn.openai.com Weight-sparse transformers have interpretable circuits [pdf] 78 points 0x79de 11 months ago 46 comments
