MVRB - Composed Screenshot Retrieval: leaderboard
MVRB composed-screenshot-retrieval slice measures retrieval with composed visual-text conditions.
Metric: Score. Source: huggingface.co. 11 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | BGE-VL-Screenshot | 59.58 | |
| 2 | VLM2Vec | 48.05 | |
| 3 | MM-Embed | 40.93 | |
| 4 | DSE | 37.78 | |
| 5 | GME | 37.68 | |
| 6 | Uni-IR | 35.92 | |
| 7 | ColPali | 35 | |
| 8 | SigLIP | 34.48 | |
| 9 | E5-V | 26.59 | |
| 10 | CLIP | 25.39 |
Interactive version: theaggregate.ai/benchmark?slug=mvrb-composed-screenshot-retrieval · How It Works · Data refreshed daily, snapshot 2026-10-20.