ImpossibleBench: Measuring Reward Hacking in LLM Coding Agents

2 pointsposted 11 hours ago
by gmays
(lesswrong.com)

No comments yet