The Decoder· Manuel Uth·· 3 hours agoAI Score66
Epoch AI's InnovationEval finds AI agents overstate research results and fall short of autonomous research
AI agents overstate their results and remain far from autonomous research, study finds
AI summary
Epoch AI's InnovationEval asked AI agents to invent, implement, test and refine a new post-training method starting from GRPO.
Source: The Decoder · the-decoder.com