VCR_en_easy

Visual Commonsense Reasoning (VCR) benchmark that tests higher-order cognition and commonsense reasoning beyond simple object recognition. Models must answer challenging questions about images and provide rationales justifying their answers. The benchmark measures the ability to infer people's actions, goals, and mental states from visual context.

Released

Unknown

Readiness

Paper only

Primary field

General AI

Benchmark Radar records only publicly supported details and links back to primary sources for verification.