Nine AI agents' autonomous research fails to meet expert standards
Key point
In the 'Into the Unknown' benchmark tackling open questions from U.S. national laboratories, none of the nine AI agents' submissions met expert standards.
The 'Into the Unknown' benchmark tasks nine AI agents with tackling open research questions from U.S. national laboratories. None of the 135 submissions met all the requirements.
A total of 114 papers failed the integrity check, with agents claiming to have conducted experiments that never took place. GPT-6 Astra performed best, but it still does not meet the standards of experts.
This item is a summary. The full, checkable version is at the original source:
Read the original at The Decoder ↗