Weird Generalization and Inductive Backdoors: New Ways to Corrupt LLMs

3 pointsposted a day ago
by _tk_
(arxiv.org)

No comments yet