Predicting Rare LLM Failures with 30× Fewer Rollouts

2 pointsposted 8 hours ago
by aranguri
(lesswrong.com)

No comments yet