# Long Video Understanding on VideoMME Long w/o sub

Source: [Long Video Understanding on VideoMME Long w/o sub](<https://www.sota2.com/research/sota/long-video-understanding-on-videomme-long-w-o-sub>)

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

Data updated: 2026-03-24T17:36:32.267Z

Evaluation of long video understanding capabilities on the VideoMME Long dataset, specifically using the subset excluding subtitles, measured by Accuracy.

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 | Long Video Understanding |
| Dataset (including any stated split) | VideoMME Long w/o sub |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 16 of 16 |
| 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

- [Accuracy](<https://www.sota2.com/research/sota/long-video-understanding-on-videomme-long-w-o-sub.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| LLaVA-Video | 70.3 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| Gemini-1.5-Pro | 67.4 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| VideoDetective (SeedVL-1.5) | 65.6 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| GPT-4o | 65.3 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| Qwen2.5-VL | 64.6 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| SeedVL-1.5 | 63.1 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| VideoDetective (Qwen3-VL) | 55.6 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| LongVITA-16k | 54.7 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| Qwen2.5-VL | 53.9 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| LongVILA | 53 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| LLaVA-Video | 52.9 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| VideoXL | 52.3 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| InternVL-2.5 | 50.8 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| Qwen3-VL | 50.2 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| VITA-1.5 | 47.1 | Not recorded | 69c21ad90636e5bf4afc4dd5 |
| LLaVA-OneVision | 46.7 | Not recorded | 69c21ad90636e5bf4afc4dd5 |

### LLaVA-Video

Accuracy: 70.3. Result ID: 69c2cb8547f34ce924acabcd.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a561e209131cea82934.md>)

[Implementation](<https://llava-vl.github.io/blog/llava-video>)

Recorded evaluation configuration: {"Param":"72B","Frames":"64"}. This record does not assert independent reproduction.

### Gemini-1.5-Pro

Accuracy: 67.4. Result ID: 69c2cb8547f34ce924acabc0.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76361e209131cea6be11.md>)

Recorded evaluation configuration: {"Param":"-","Frames":"256"}. This record does not assert independent reproduction.

### VideoDetective (SeedVL-1.5)

Accuracy: 65.6. Result ID: 69c2cb8547f34ce924acabce.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Implementation](<https://github.com/yangruoliu/VideoDetective>)

Recorded evaluation configuration: {"Param":"20B(A)","Frames":"32"}. This record does not assert independent reproduction.

### GPT-4o

Accuracy: 65.3. Result ID: 69c2cb8547f34ce924acabbf.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75681e209131cea67a98.md>)

Recorded evaluation configuration: {"Param":"-","Frames":"384"}. This record does not assert independent reproduction.

### Qwen2.5-VL

Accuracy: 64.6. Result ID: 69c2cb8547f34ce924acabcc.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Param":"72B","Frames":"128"}. This record does not assert independent reproduction.

### SeedVL-1.5

Accuracy: 63.1. Result ID: 69c2cb8547f34ce924acabc1.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78791e209131cea788a1.md>)

Recorded evaluation configuration: {"Param":"20B(A)","Frames":"32"}. This record does not assert independent reproduction.

### VideoDetective (Qwen3-VL)

Accuracy: 55.6. Result ID: 69c2cb8547f34ce924acabcb.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Implementation](<https://github.com/yangruoliu/VideoDetective>)

Recorded evaluation configuration: {"Param":"8B","Frames":"32"}. This record does not assert independent reproduction.

### LongVITA-16k

Accuracy: 54.7. Result ID: 69c2cb8547f34ce924acabc2.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa770d1e209131cea70cc6.md>)

Recorded evaluation configuration: {"Param":"14B","Frames":"64"}. This record does not assert independent reproduction.

### Qwen2.5-VL

Accuracy: 53.9. Result ID: 69c2cb8547f34ce924acabc7.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71d3c.md>)

Recorded evaluation configuration: {"Param":"7B","Frames":"128"}. This record does not assert independent reproduction.

### LongVILA

Accuracy: 53. Result ID: 69c2cb8547f34ce924acabc3.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76271e209131cea6b860.md>)

[Implementation](<https://www.sota2.com/NVlabs/LongVILA>)

Recorded evaluation configuration: {"Param":"7B","Frames":"1fps"}. This record does not assert independent reproduction.

### LLaVA-Video

Accuracy: 52.9. Result ID: 69c2cb8547f34ce924acabc5.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a561e209131cea82934.md>)

[Implementation](<https://llava-vl.github.io/blog/llava-video>)

Recorded evaluation configuration: {"Param":"7B","Frames":"512"}. This record does not assert independent reproduction.

### VideoXL

Accuracy: 52.3. Result ID: 69c2cb8547f34ce924acabc6.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa762a1e209131cea6b9cb.md>)

Recorded evaluation configuration: {"Param":"7B","Frames":"1fps"}. This record does not assert independent reproduction.

### InternVL-2.5

Accuracy: 50.8. Result ID: 69c2cb8547f34ce924acabc9.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bab1e209131cea89c7f.md>)

[Implementation](<https://github.com/OpenGVLab/InternVL>)

Recorded evaluation configuration: {"Param":"8B","Frames":"32"}. This record does not assert independent reproduction.

### Qwen3-VL

Accuracy: 50.2. Result ID: 69c2cb8547f34ce924acabc8.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6927db54d6b0dd3342930f66.md>)

Recorded evaluation configuration: {"Param":"8B","Frames":"32"}. This record does not assert independent reproduction.

### VITA-1.5

Accuracy: 47.1. Result ID: 69c2cb8547f34ce924acabca.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76ae1e209131cea6e684.md>)

Recorded evaluation configuration: {"Param":"7B","Frames":"16"}. This record does not assert independent reproduction.

### LLaVA-OneVision

Accuracy: 46.7. Result ID: 69c2cb8547f34ce924acabc4.

[Reporting paper](<https://www.sota2.com/research/paper/69c21ad90636e5bf4afc4dd5.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa756f1e209131cea67e58.md>)

[Implementation](<https://llava-vl.github.io/blog/llava-onevision>)

Recorded evaluation configuration: {"Param":"7B","Frames":"-"}. This record does not assert independent reproduction.

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