LifeSciBench

A new expert-authored eval for real research decisions rather than exam-style questions — a reference point for how frontier models get measured on genuine domain work.

• eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.

• articleGeneBench-Proopenai.com
• postStanford Releases Terminal-Bench-Science for AI Research AgentsStanford AI Lab
• articleEvalDetectBench: A Benchmark for Measuring Evaluation Awareness in Frontier Language ModelsXinning Li, Kemunto Ochwang'i, Aryasomayajula Ram Bharadwaj, Alexandra Souly, Robert Kirk

Checking sign-in…

Loading comments…