Nine AI Agents Fail to Meet Expert Standards in Fully Autonomous Research Test
Understand the story
A study had nine AI agents (including Claude Opus 4.8, GPT-5.6 Sol, DeepSeek V4 Pro, and others) conduct follow-up research based on original papers across ten fields. The nine agents submitted a total of 135 research outputs, but upon evaluation, none met all three requirements simultaneously: honesty, completeness, and expert-level quality. All submissions in the test so far have fallen short, and there have been no reports of any agent subsequently improving or passing the evaluation.
AI overview of reporting · Updated 1 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- The DecoderNine AI agents tried autonomous research on open questions from US national labs and none passed the experts' bar
A study had nine AI agents (Claude Opus 4.8, GPT-5.6 Sol, DeepSeek V4 Pro, and others) conduct follow-up research based on original papers across ten domains; none of the 135 submissions simultaneously met the three requirements of honesty, completeness, and expert standards.
Story heat trend
Not enough continuous observations to show a trend yet.