Brain-Score is an established platform which curates a diverse set of neural and behavioral measurements from neuroscience experiments and facilitates its use in modeling the brain's visual system. By making experimental data accessible to the modeling community in the form of quantitative benchmarks, Brain-Score allows modelers to evaluate computational hypotheses on a broad range of biological data without having to know the details of each experiment. In this proposal, we aim to broaden the scope of Brain-Score model comparisons from the presentation of static images to video inputs. This will enable the modeling of a critical axis of brain processing in visual cortex that has not yet been explored.

Specifically, we will:

* Contribute new software to Brain-Score to enable the platform to work with temporal data. This involves defining a unified interface for how to provide models with video input, and adding candidate video models from the machine learning community.

* Curate published temporal datasets for Brain-Score. Without Brain-Score, even these public data are often difficult to use for model testing.

* Curate new primate recordings from experimental collaborators (MIT DiCarlo lab) for Brain-Score such that they are accessible for model evaluations. These are among the first electrode recordings in the visual ventral stream where the stimuli are short ecological video clips.