Multimodal Embedding on MMEB V1
69.7Classification AccuracyThink-Then-Embed
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Think-Then-EmbedBackbone=Qwen2VL-2B, Method Category=Explicit Reasoning UME2026.06 | 69.7 | 72.4 | 74 | 90.6 | 80.5 | 67 | 74.5 | |
| Think-When-NeededBackbone=Qwen3VL-4B, Method Category=Explicit Reasoning UME2026.06 | 68.6 | 71.7 | 68 | 87 | 75.4 | 66.2 | 71.3 | |
| UNITE#Params=7B, Backbone=Qwen2-VL, Pre-training approach=Contrastive Learning2025.11 | 68.3 | 65.1 | 71.6 | 84.8 | 73.6 | 66.3 | 70.3 | |
| Ops-MM-EmbedBackbone=Qwen2VL-2B, Method Category=Discriminative UME2026.06 | 68.1 | 65.1 | 69.2 | 80.9 | 71.8 | 65.6 | 69 | |
| RIMEBackbone=Qwen2VL-2B, Method Category=Explicit Reasoning UME2026.06 | 67.9 | 64.4 | 69.8 | 82.1 | 64.5 | 72.9 | 69.2 | |
| mmE5#Params=11B, Backbone=Llama-3.2-Vision, Pre-training approach=Contrastive Learning2025.11 | 67.6 | 62.8 | 70.9 | 89.7 | 72.3 | 66.7 | 69.8 | |
| LaMEBackbone=Qwen2VL-2B, Method Category=Latent Reasoning UME2026.06 | 67.6 | 66.2 | 70.5 | 81.2 | 73.1 | 64.4 | 69.3 | |
| CoMa#Params=7B, Backbone=Qwen2.5-VL, Pre-training approach=Additional pre-training2025.11 | 67.4 | 70.6 | 72.4 | 87.6 | 75.2 | 67.6 | 72.2 | |
| PLUMEBackbone=Qwen2VL-2B, Method Category=Latent Reasoning UME2026.06 | 66.1 | 59.5 | 67.6 | 79.9 | 69.6 | 62.2 | 66.3 | |
| MoCa#Params=7B, Backbone=Qwen2.5-VL, Pre-training approach=Additional pre-training2025.11 | 65.8 | 64.7 | 75 | 92.4 | 74.7 | 67.6 | 71.5 | |
| ReMatchBackbone=Qwen2VL-2B, Method Category=Discriminative UME2026.06 | 65.8 | 65.9 | 70.1 | 83.3 | 72.8 | 64.7 | 69.2 | |
| CAFe#Params=7B, Backbone=LLaVA-OV, Pre-training approach=Contrastive Learning2025.11 | 65.2 | 65.6 | 70 | 91.2 | 75.8 | 62.4 | 69.8 | |
| UME-R1Backbone=Qwen2VL-2B, Method Category=Explicit Reasoning UME2026.06 | 64.8 | 62.8 | 67.6 | 77.2 | 71.5 | 60.4 | 66.6 | |
| RGEBackbone=Qwen3VL-2B, Method Category=Explicit Reasoning UME2026.06 | 64.4 | 67.8 | 70.2 | 90.1 | 75.3 | 63.7 | 70.1 | |
| UNITE#Params=2B, Backbone=Qwen2-VL, Pre-training approach=Contrastive Learning2025.11 | 63.2 | 55.9 | 65.4 | 75.6 | 65.8 | 60.1 | 63.3 | |
| VLM2Vec-V2Backbone=Qwen2VL-2B, Method Category=Discriminative UME2026.06 | 62.9 | 56.3 | 69.5 | 77.3 | 68.8 | 60.1 | 64.9 | |
| Embed-RLBackbone=Qwen2.5VL-3B, Method Category=Explicit Reasoning UME2026.06 | 62.8 | 67.9 | 68.6 | 90.4 | 65.9 | 71.9 | 69.2 | |
| VLM2Vec#Params=7B, Backbone=Qwen2-VL, Pre-training approach=Contrastive Learning2025.11 | 62.6 | 57.8 | 69.9 | 81.7 | 72.2 | 57.8 | 65.8 | |
| CoMa#Params=3B, Backbone=Qwen2.5-VL, Pre-training approach=Additional pre-training2025.11 | 61.3 | 65.1 | 70 | 82.7 | 71.3 | 61.6 | 67.5 | |
| UniME#Params=7B, Backbone=LLaVA-1.6, Pre-training approach=Additional pre-training2025.11 | 60.6 | 52.9 | 67.9 | 85.1 | 68.4 | 57.9 | 66.6 | |
| MoCa#Params=3B, Backbone=Qwen2.5-VL, Pre-training approach=Additional pre-training2025.11 | 59.8 | 62.9 | 70.6 | 88.6 | 72.3 | 61.5 | 67.5 | |
| VLM2VecBackbone=Qwen2VL-2B, Method Category=Discriminative UME2026.06 | 58.7 | 49.3 | 65 | 72.9 | 64.8 | 53.4 | 59.7 | |
| GME#Params=2B, Backbone=Qwen2-VL, Pre-training approach=Contrastive Learning2025.11 | 56.9 | 41.2 | 67.8 | 53.4 | — | — | — | |
| MMRet#Params=7B, Backbone=LLaVA-1.6, Pre-training approach=Contrastive Learning2025.11 | 56 | 57.4 | 69.9 | 83.6 | 68 | 59.1 | 65.8 | |
| VLM2Vec#Params=3B, Backbone=Qwen2.5-VL, Pre-training approach=Contrastive Learning2025.11 | 55.3 | 57.3 | 62.7 | 73.2 | — | — | 60.3 | |
| CLIP#Params=0.4B, Backbone=ViT-L, Pre-training approach=Contrastive Learning2025.11 | 55.2 | 19.7 | 53.2 | 62.2 | 47.6 | 42.8 | 45.4 | |
| UniME#Params=4.2B, Backbone=Phi3.5-V, Pre-training approach=Additional pre-training2025.11 | 54.8 | 55.9 | 64.5 | 81.8 | 68.2 | 52.7 | 64.2 | |
| GMEBackbone=Qwen2VL-2B, Method Category=Discriminative UME2026.06 | 54.4 | 29.9 | 66.9 | 55.5 | 49.2 | 55.2 | 51.9 | |
| CLIPBackbone=CLIP-L-0.4B, Method Category=Discriminative UME2026.06 | 42.8 | 9.1 | 53 | 51.8 | 37.1 | 38.7 | 37.8 | |
| OpenCLIP#Params=0.4B, Backbone=ViT-L, Pre-training approach=Contrastive Learning2025.11 | 41.5 | 6.9 | 44.6 | 53.5 | 32.8 | 36 | 36.6 | |
| E5-V#Params=7B, Backbone=LLaVA-1.6, Pre-training approach=Contrastive Learning2025.11 | 39.7 | 10.8 | 39.4 | 60.2 | 34.2 | 33.9 | 33.9 |