Tag

#neurips-2026

1 story tagged neurips-2026.

Radar chart comparing two AI-agent-generated papers across six review criteria normalized to percentage of maximum. Paper 1 scored 50% quality, 25% clarity, 50% significance, 75% originality, 33% overall, 80% confidence. Paper 2 scored 25% quality, 50% clarity, 50% significance, 50% originality, 17% overall, 100% confidence. Both papers were rejected. Primary keyword: shadow evaluation.

Shadow evaluation shows AI agents fail open-ended research

Shadow evaluation tests whether AI agents can do open-ended research. They wrote code and ran experiments, but both papers were rejected at 2/6 and 1/6.

Lars Cornelissen · Aug 18, 2026