Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMs

3 pointsposted 2 months ago
by _tk_
(arxiv.org)

No comments yet