VISTA (Video Grounding) - Object-Object: leaderboard

VISTA (2026) spatio-temporal video grounding on queries describing interactions between objects. Mean spatio-temporal IoU in percent; higher is better.

Metric: Mean spatio-temporal IoU (m_vIoU, %): per-frame box IoU summed over the frames where predicted and true tubes overlap in time, divided by their temporal union, on object-object interactions, over the 11,814 video-query pairs of VISTA (HCSTVG-v1 and v2, VidVRD, VidSTG, MeViS and RVOS videos re-annotated with an interaction taxonomy), zero-shot on sub-sampled frames; the model must localize the queried subject with a box in every frame; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Qwen 3 VL 8B59.1#917
2VISTA CogVLM Grounding (checkpoint unspecified)54
3InternVL2.5-8B47.7#2370
4MiniGPT-v243.1
5VISTA Sphinx-v2 (checkpoint unspecified)42.5
6Qwen-VL-Chat31.7#7860
7LLaVA-Grounding-7B28.4
8VISTA MiMo-VL-7B (checkpoint unspecified)27
9Shikra-7B24.6
10Ferret-7B13.4

Interactive version: theaggregate.ai/benchmark?slug=vista-video-grounding-object-object · How It Works · Data refreshed daily, snapshot 2026-11-21.