SciResearch-Eval finds scientific agents overconfident when simulators omit physics
On 93 tasks, most models' 95% intervals miss the truth, while GPT-6 Astra and Claude Fable 5.2 stay near calibrated.
1 stories, newest first · one card per story
On 93 tasks, most models' 95% intervals miss the truth, while GPT-6 Astra and Claude Fable 5.2 stay near calibrated.
Today 986 readers · 1.2K views · 166 reads · all time 202K views