# Multimodal Dense Retrieval on MMEB V2

Source: [Multimodal Dense Retrieval on MMEB V2](<https://www.sota2.com/research/sota/multimodal-dense-retrieval-on-mmeb-v2>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-26T06:47:52.187Z

Evaluation of model performance on the MMEB V2 benchmark suite, specifically focusing on Multimodal Dense Retrieval tasks across image and video modalities, reporting detailed metrics like Recall@1 and NDCG@5.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Multimodal Dense Retrieval |
| Dataset (including any stated split) | MMEB V2 |
| Metric | Image Recall@1 |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 10 of 10 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [Image Recall@1](<https://www.sota2.com/research/sota/multimodal-dense-retrieval-on-mmeb-v2.md?metric=Image%20Recall%401>): Higher is better; unit: PERCENT.

- [Visdoc VDRv1 NDCG@5](<https://www.sota2.com/research/sota/multimodal-dense-retrieval-on-mmeb-v2.md?metric=Visdoc%20VDRv1%20NDCG%405>): Higher is better; unit: PERCENT.

- [Visdoc VDRv2 NDCG@5](<https://www.sota2.com/research/sota/multimodal-dense-retrieval-on-mmeb-v2.md?metric=Visdoc%20VDRv2%20NDCG%405>): Higher is better; unit: PERCENT.

- [Visdoc VR NDCG@5](<https://www.sota2.com/research/sota/multimodal-dense-retrieval-on-mmeb-v2.md?metric=Visdoc%20VR%20NDCG%405>): Higher is better; unit: PERCENT.

- [Visdoc OOD NDCG@5](<https://www.sota2.com/research/sota/multimodal-dense-retrieval-on-mmeb-v2.md?metric=Visdoc%20OOD%20NDCG%405>): Higher is better; unit: PERCENT.

- [Video Recall@1](<https://www.sota2.com/research/sota/multimodal-dense-retrieval-on-mmeb-v2.md?metric=Video%20Recall%401>): Higher is better; unit: PERCENT.

- [Average Score](<https://www.sota2.com/research/sota/multimodal-dense-retrieval-on-mmeb-v2.md?metric=Average%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Image Recall@1 | Result date | Report paper |
| --- | --- | --- | --- |
| Qwen3-VL-Embed-8B | 80.15 | Not recorded | 6a152916dbd7150506f7ded0 |
| Qwen3-VL-Embed-8B | 80.09 | Not recorded | 6a152916dbd7150506f7ded0 |
| Qwen3-VL-Embed-2B | 75.33 | Not recorded | 6a152916dbd7150506f7ded0 |
| Qwen3-VL-Embed-2B | 74.91 | Not recorded | 6a152916dbd7150506f7ded0 |
| GME-7B | 73.57 | Not recorded | 6a152916dbd7150506f7ded0 |
| GME-7B | 73.09 | Not recorded | 6a152916dbd7150506f7ded0 |
| GME-2B | 71.05 | Not recorded | 6a152916dbd7150506f7ded0 |
| GME-2B | 70.13 | Not recorded | 6a152916dbd7150506f7ded0 |
| VLM2Vec-V2.0 | 69.95 | Not recorded | 6a152916dbd7150506f7ded0 |
| VLM2Vec-V2.0 | 69.5 | Not recorded | 6a152916dbd7150506f7ded0 |

### Qwen3-VL-Embed-8B

Image Recall@1: 80.15. Result ID: 6a15418cfbc373592c0c0286.

[Reporting paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Implementation](<https://github.com/HanSolo9682/SMART>)

Recorded evaluation configuration: {"SMART":"true"}. This record does not assert independent reproduction.

### Qwen3-VL-Embed-8B

Image Recall@1: 80.09. Result ID: 6a15418cfbc373592c0c0285.

[Reporting paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Implementation](<https://github.com/HanSolo9682/SMART>)

Recorded evaluation configuration: {"SMART":"false"}. This record does not assert independent reproduction.

### Qwen3-VL-Embed-2B

Image Recall@1: 75.33. Result ID: 6a15418cfbc373592c0c0284.

[Reporting paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Implementation](<https://github.com/HanSolo9682/SMART>)

Recorded evaluation configuration: {"SMART":"true"}. This record does not assert independent reproduction.

### Qwen3-VL-Embed-2B

Image Recall@1: 74.91. Result ID: 6a15418cfbc373592c0c0283.

[Reporting paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Implementation](<https://github.com/HanSolo9682/SMART>)

Recorded evaluation configuration: {"SMART":"false"}. This record does not assert independent reproduction.

### GME-7B

Image Recall@1: 73.57. Result ID: 6a15418cfbc373592c0c0282.

[Reporting paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Implementation](<https://github.com/HanSolo9682/SMART>)

Recorded evaluation configuration: {"SMART":"true"}. This record does not assert independent reproduction.

### GME-7B

Image Recall@1: 73.09. Result ID: 6a15418cfbc373592c0c0281.

[Reporting paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Implementation](<https://github.com/HanSolo9682/SMART>)

Recorded evaluation configuration: {"SMART":"false"}. This record does not assert independent reproduction.

### GME-2B

Image Recall@1: 71.05. Result ID: 6a15418cfbc373592c0c0280.

[Reporting paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Implementation](<https://github.com/HanSolo9682/SMART>)

Recorded evaluation configuration: {"SMART":"true"}. This record does not assert independent reproduction.

### GME-2B

Image Recall@1: 70.13. Result ID: 6a15418cfbc373592c0c027f.

[Reporting paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Implementation](<https://github.com/HanSolo9682/SMART>)

Recorded evaluation configuration: {"SMART":"false"}. This record does not assert independent reproduction.

### VLM2Vec-V2.0

Image Recall@1: 69.95. Result ID: 6a15418cfbc373592c0c027e.

[Reporting paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Implementation](<https://github.com/HanSolo9682/SMART>)

Recorded evaluation configuration: {"SMART":"true"}. This record does not assert independent reproduction.

### VLM2Vec-V2.0

Image Recall@1: 69.5. Result ID: 6a15418cfbc373592c0c027d.

[Reporting paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a152916dbd7150506f7ded0.md>)

[Implementation](<https://github.com/HanSolo9682/SMART>)

Recorded evaluation configuration: {"SMART":"false"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
