AI Agents Fail Open-Ended Research: Authors Reject Papers Generated with 6 Days & Thousands in Compute
sayashk · x · 2026-07-31
Sayash Kapoor and colleagues evaluated whether AI agents can conduct open-ended AI research, introducing a method called CRUXes (shadow evaluations). Agents were given research questions from unpublished papers, six days, and thousands of dollars in compute. The original authors then reviewed the AI-generated outputs.
The authors unambiguously rejected the agents' papers. While fluent in engineering tasks, the agents struggled with open-ended research requiring hypothesis generation, evidence selection, and recognizing failing approaches. However, researchers note that expanding these evaluations to different types of papers might soon reveal where models eventually succeed.
More from AGI Musings
- Miles Cranmer's talk: how to avoid 'AI Brain Rot' and intelligence atrophy — Scobleizer · 2026-10-16
- AI's fundamental design flaws are fully hidden by its products' superficial design flaws — lescarr · 2026-10-16
- IMF Official: Markets' Surprising Resilience Largely Rests on AI Optimism — kernelangus420 · 2026-10-16
- Can AI trained only on human content infer the shape of the human brain? — rickasaurus · 2026-10-16
- Vals RSI Index extrapolates Google models hitting 130% autonomous AI R&D by Oct 2027 — HaydnBelfield · 2026-10-16
- Gowers' tl;dr: no p(doom), but can't confidently say it's close to zero — michael_nielsen · 2026-10-16