OPERA: A Unified Omnimodal Progressive Spatio-Temporal Reasoning Agent for Referring Video Segmentation
A unified agent for referring video segmentation with text, audio, and reference-image queries through progressive temporal and spatial reasoning.