We have compiled a set of datasets that are helpful for benchmarking models for computational behavior. Studying model performance on these open-source datasets can help standardize evaluation procedures, understand gaps in model performance, and improve reproducibility. Click on an entry in the table to see more information about the data and download links.
This list is not exhaustive and we would appreciate your input on any other useful datasets that we might have missed!
Other notable repos:
Papers With Code, 5000+ datasets for general machine learning tasks. https://paperswithcode.com/datasets
OpenBehavior, behavioral videos donated from a variety of labs. https://edspace.american.edu/openbehavior/video-repository/video-repository-2/
Maintainers for Computational Behavior Data Repo: Jennifer J. Sun (Caltech), Ann Kennedy (Northwestern)
Contributors: ... maybe you?