Temporal Grounding
A task requiring a model to output the time interval in which a referenced target appears in a video, typically expressed as start and end timestamps. Evaluation commonly uses IoU or mIoU to measure whether the model can anchor a language description to the correct segment on the timeline.
Core Facts
Timeline (last 90 days)
Each track of OmniCompass contains three types of complementary tasks: temporal grounding (TG), spatial grounding (SG), and question answering (QA).
Even on leading models, precise temporal and spatial localization remains a core unsolved challenge; models perform acceptably on coarse-grained semantic understanding, but have clear gaps on precise localization tasks.
Multimodal understanding of video content requires the integrated use of capabilities such as speech-to-text, scene recognition, and object detection.
All Facts (3)
Multimodal understanding of video content requires the integrated use of capabilities such as speech-to-text, scene recognition, and object detection.
75%UnverifiedEach track of OmniCompass contains three types of complementary tasks: temporal grounding (TG), spatial grounding (SG), and question answering (QA).
50%UnverifiedEven on leading models, precise temporal and spatial localization remains a core unsolved challenge; models perform acceptably on coarse-grained semantic understanding, but have clear gaps on precise localization tasks.
50%