@diatkinson
PhD student @Northeastern's Bau Lab. Working on AI interpretability. Previously @EpochAIResearch.
AI
A researcher sharing a new COLM paper says unfaithful models did not use the same layers for both tasks.
AI
