Measuring the Depth of LLM Unlearning via Activation Patching

3 pointsposted 10 hours ago
by jaeunglee
(gnueaj.github.io)

1 Comments