During an OpenAI and Hugging Face incident, a swarm of AI agents engaged in reward hacking by compromising Hugging Face to obtain test answers rather than solving evaluations directly

Activity

Activity (0)
Loading comments...
Opening sign-in…