# Multimodal Retrieval on MMEB Video V2

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

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

Data updated: 2026-03-04T12:15:08.554Z

Evaluation of model performance on Multimodal Retrieval tasks specifically within the MMEB Video V2 benchmark suite, reporting scores for Classification (CLS), Question Answering (QA), Retrieval (RET), and Multimodal Retrieval (MRET).

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 Retrieval |
| Dataset (including any stated split) | MMEB Video V2 |
| Metric | CLS Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| 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

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

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

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

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

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

## Recorded results

| Method | CLS Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| UME-R1 | 51.6 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| UME-R1 | 48.6 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| UME-R1 | 47 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| UME-R1 | 44.3 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| LamRA | 39.3 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| VLM2Vec-V2 | 39.3 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| VLM2Vec | 39.1 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| DUME | 37.7 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| GME | 37.4 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| CAFe | 35.8 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| GME | 34.9 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| VLM2Vec | 33.4 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| LamRA | 32.9 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| DUME | 32.9 | Not recorded | 6909887c3ef7ccd74e5e5b46 |
| ColPali-V1.3 | 26.7 | Not recorded | 6909887c3ef7ccd74e5e5b46 |

### UME-R1

CLS Accuracy: 51.6. Result ID: 69a822386dbb7075b94e34a6.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Implementation](<https://github.com/XMUDeepLIT/UME-R1>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"7B","Oracle Evaluation Mode":"true"}. This record does not assert independent reproduction.

### UME-R1

CLS Accuracy: 48.6. Result ID: 69a822386dbb7075b94e34a4.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Implementation](<https://github.com/XMUDeepLIT/UME-R1>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"7B"}. This record does not assert independent reproduction.

### UME-R1

CLS Accuracy: 47. Result ID: 69a822386dbb7075b94e34a5.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Implementation](<https://github.com/XMUDeepLIT/UME-R1>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"2B","Oracle Evaluation Mode":"true"}. This record does not assert independent reproduction.

### UME-R1

CLS Accuracy: 44.3. Result ID: 69a822386dbb7075b94e34a3.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Implementation](<https://github.com/XMUDeepLIT/UME-R1>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"2B"}. This record does not assert independent reproduction.

### LamRA

CLS Accuracy: 39.3. Result ID: 69a822386dbb7075b94e349b.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76081e209131cea6b06a.md>)

[Implementation](<https://code-kunkun.github.io/LamRA/>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"7B"}. This record does not assert independent reproduction.

### VLM2Vec-V2

CLS Accuracy: 39.3. Result ID: 69a822386dbb7075b94e349f.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79d61e209131cea7fa13.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"2B"}. This record does not assert independent reproduction.

### VLM2Vec

CLS Accuracy: 39.1. Result ID: 69a822386dbb7075b94e349e.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa766b1e209131cea6cd6f.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"7B"}. This record does not assert independent reproduction.

### DUME

CLS Accuracy: 37.7. Result ID: 69a822386dbb7075b94e34a1.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"2B"}. This record does not assert independent reproduction.

### GME

CLS Accuracy: 37.4. Result ID: 69a822386dbb7075b94e349a.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"7B"}. This record does not assert independent reproduction.

### CAFe

CLS Accuracy: 35.8. Result ID: 69a822386dbb7075b94e34a0.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa779f1e209131cea746ef.md>)

Recorded evaluation configuration: {"Backbone":"LLaVA-OV","Parameter Scale":"7B"}. This record does not assert independent reproduction.

### GME

CLS Accuracy: 34.9. Result ID: 69a822386dbb7075b94e3499.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"2B"}. This record does not assert independent reproduction.

### VLM2Vec

CLS Accuracy: 33.4. Result ID: 69a822386dbb7075b94e349d.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa766b1e209131cea6cd6f.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"2B"}. This record does not assert independent reproduction.

### LamRA

CLS Accuracy: 32.9. Result ID: 69a822386dbb7075b94e349c.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76081e209131cea6b06a.md>)

[Implementation](<https://code-kunkun.github.io/LamRA/>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL","Parameter Scale":"7B"}. This record does not assert independent reproduction.

### DUME

CLS Accuracy: 32.9. Result ID: 69a822386dbb7075b94e34a2.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL","Parameter Scale":"7B"}. This record does not assert independent reproduction.

### ColPali-V1.3

CLS Accuracy: 26.7. Result ID: 69a822386dbb7075b94e3498.

[Reporting paper](<https://www.sota2.com/research/paper/6909887c3ef7ccd74e5e5b46.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77261e209131cea7161f.md>)

Recorded evaluation configuration: {"Backbone":"PaliGemma","Parameter Scale":"3B"}. This record does not assert independent reproduction.

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