Function vectors as a model diffing tool: 17 heads repair a bad fine-tune
Source ↗
👁 0
💬 0
I take two fine-tuned models trained to give bad medical advice, one on Qwen2.5-7B and one on Llama-3.1-8B, from the Model Organisms for Emergent Misalignment collection, and I found out that I could make one safe by simply copying 17 attention heads from the base model it was trained from (which I'm assuming is the good one). The interesting thing is that the reverse is not true. There is also a single direction you can pull out of the difference between the two models, and removing it partly c
Comments (0)