Data coverage
BrainWideBench is a large-scale, multi-task benchmark for evaluating neural foundation models, derived from the International Brain Laboratory (IBL) Brainwide Map dataset. It contains preprocessed Neuropixels recordings from 139 mice spanning 276 brain regions during a standardized visual decision-making task. The benchmark is organized around three complementary task suites: behavioral decoding (TS1), neural activity prediction (TS2), and brain region identification (TS3). Data are processed into a standardized format using the torchbrain library and include spike-sorted neural recordings, behavioral signals (wheel movements, paw kinematics, whisker motion energy, lick detection), trial structure, stimulus parameters, and anatomical annotations. A strict animal-level train/evaluation split separates 126 pretraining animals (423 sessions) from 13 held-out evaluation animals (29 sessions), enabling reproducible evaluation of across-animal generalization.
Access & permitted use
Review the publisher’s documentation for availability, eligibility and permitted uses, including commercial analysis, model training and redistribution.
CC-BY 4.0
Need help sourcing data?
Tell us your required fields, population, geography and time period. We can review suitable sources and licensing options.
Request sourcing help