Predicted Cortex Is Not a Domain-General Prior: A Matched-Control Audit of Brain-Encoding Features for Video Memorability

Derived feature arrays for the paper It Depends on the Dataset: When a Brain-Encoding Model's Predicted Responses Beat Their Visual Backbone for Video Memorability (arXiv:2607.16292). Four NumPy arrays: TRIBE v2 predicted average-subject cortical responses (fsaverage5) and V-JEPA2 backbone embeddings for the VideoMem (820) and Memento10k (499) clips, each stored as stim_ids plus float32 X features. The dataset videos and memorability scores are not included; obtain them from the dataset owners (VideoMem: InterDigital; Memento10k: MIT) under their own licences and align by stim_ids. Shared with InterDigital's written approval for research use. Code: https://github.com/rodriguescarson/brain-encoding-video-memorability