MMEB: leaderboard

36 multimodal embedding datasets across classification, VQA, retrieval and visual grounding (20 training, 16 held-out) from TIGER-Lab at Waterloo (2024); Precision@1; V2 extends to 78 tasks.

Source: huggingface.co.

Interactive version: theaggregate.ai/benchmark?slug=mmeb · How It Works · Data refreshed daily, snapshot 2026-09-05.