We develop hypothesis tests for fairness metrics for small and large subgroups

STSBench is a benchmark that evaluates the capabilities of Multi-modal Large Language Models to reason about spatio-temporal actions.