The traditional paradigm of AI benchmarking has long relied on static, closed-ended datasets where the correct answer is known in advance.
Into the Unknown: Benchmarking AI agents on open scientific research problems
calendar_today
October 15, 2026
domain
aaaai-me