NeuroscienceBench debuts: 91 evals test if AI agents can analyze real neuroscience data
kenbwork · x · 2026-11-10
A new benchmark series kicks off with NeuroscienceBench, testing whether AI agents can analyze experimental data and draw reliable biological conclusions, with immunology and aging evals to follow.
- Composition: 91 evaluations drawn from 43 peer-reviewed studies, spanning sequencing, electrophysiology, EEG, brain-computer interface decoding, imaging, behavior, and clinical biomarkers.
- Format: agents analyze real experimental data and return structured answers graded deterministically.
- Results (chart): pass rates across 21 models are low overall (mean 20%); Opus 5.5 leads bulk RNA-seq at 50%, while patch-clamp/MEA imaging tasks sit below 20% for every model, underscoring a wide gap in scientific data analysis.
More from Models
- Claude Fable 5.2 and Mythos 5.2 spotted on AWS Bedrock, priced 20% below 5.1 — testingcatalog · 2026-11-10
- Meta's Muse Ember finds $2.48M misallocation across 13 fund documents in Box — alexandr_wang · 2026-11-10
- MiniMax H3 video generation running locally on just 6GB VRAM — No_Cow26 · 2026-11-10
- If Meta squeezed that much intelligence per compute into a small model, larger ones should follow — daniel_mac8 · 2026-11-10
- Ripple-65B-Preview tops open-weight agentic board with 98.2% job completion — ryanshrout · 2026-11-10
- Meta's Ember tops Vals Index on cost at $0.15 per test, fastest in top 12 — teortaxesTex · 2026-11-10