# Video Question Answering on OVO-Bench Backward Tracing

Source: [Video Question Answering on OVO-Bench Backward Tracing](<https://www.sota2.com/research/sota/video-question-answering-on-ovo-bench-backward-tracing>)

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

Data updated: 2026-05-11T10:39:53.434Z

Evaluation of model performance on the OVO-Bench Backward Tracing dataset for Video Question Answering, reporting metrics including EPM, ASI, HLD, and Average.

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 | Video Question Answering |
| Dataset (including any stated split) | OVO-Bench Backward Tracing |
| Metric | EPM |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 17 of 17 |
| 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

- [EPM](<https://www.sota2.com/research/sota/video-question-answering-on-ovo-bench-backward-tracing.md?metric=EPM>): Higher is better; unit: PERCENT.

- [ASI](<https://www.sota2.com/research/sota/video-question-answering-on-ovo-bench-backward-tracing.md?metric=ASI>): Higher is better; unit: PERCENT.

- [HLD](<https://www.sota2.com/research/sota/video-question-answering-on-ovo-bench-backward-tracing.md?metric=HLD>): Higher is better; unit: PERCENT.

- [Average Score](<https://www.sota2.com/research/sota/video-question-answering-on-ovo-bench-backward-tracing.md?metric=Average%20Score>): Higher is better; unit: PERCENT.

## Recorded results

| Method | EPM | Result date | Report paper |
| --- | --- | --- | --- |
| ViSpeak | 59.93 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| StreamForest | 58.92 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| Gemini 1.5 Pro | 58.59 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| GPT-4o | 57.91 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| TimeChat-Online | 55.9 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| ThinkStream | 53.87 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| Qwen2.5-VL-7B | 51.52 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| LLaVA-Video | 51.18 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| Qwen2.5-VL-7B + SAVEMem | 50.84 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| Qwen2.5-VL-3B | 49.83 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| Dispider | 48.48 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| Qwen2.5-VL-3B + SAVEMem | 47.81 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| Qwen2-VL | 44.44 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| InternVL2 | 43.1 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| LongVU | 43.1 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| Flash-VStream | 36.36 | Not recorded | 6a0162c2dbc836b87d8233d0 |
| VideoLLM-Online | 22.22 | Not recorded | 6a0162c2dbc836b87d8233d0 |

### ViSpeak

EPM: 59.93. Result ID: 6a01b08f47f34ce924e44383.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77741e209131cea733a6.md>)

Recorded evaluation configuration: {"Size":"7B","Frames":"1 fps","Mode":"Online","Training Strategy":"Training-Based"}. This record does not assert independent reproduction.

### StreamForest

EPM: 58.92. Result ID: 6a01b08f47f34ce924e44386.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bae1e209131cea89d70.md>)

Recorded evaluation configuration: {"Size":"7B","Frames":"1 fps","Mode":"Online","Training Strategy":"Training-Based"}. This record does not assert independent reproduction.

### Gemini 1.5 Pro

EPM: 58.59. Result ID: 6a01b08f47f34ce924e4437a.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

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

Recorded evaluation configuration: {"Size":"-","Frames":"1 fps","Mode":"Proprietary"}. This record does not assert independent reproduction.

### GPT-4o

EPM: 57.91. Result ID: 6a01b08f47f34ce924e4437b.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

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

Recorded evaluation configuration: {"Size":"-","Frames":"64","Mode":"Proprietary"}. This record does not assert independent reproduction.

### TimeChat-Online

EPM: 55.9. Result ID: 6a01b08f47f34ce924e44385.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78311e209131cea771c4.md>)

Recorded evaluation configuration: {"Size":"7B","Frames":"1 fps","Mode":"Online","Training Strategy":"Training-Based"}. This record does not assert independent reproduction.

### ThinkStream

EPM: 53.87. Result ID: 6a01b08f47f34ce924e44384.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b78e730636e5bf4afbdb5e.md>)

Recorded evaluation configuration: {"Size":"3B","Frames":"1 fps","Mode":"Online","Training Strategy":"Training-Based"}. This record does not assert independent reproduction.

### Qwen2.5-VL-7B

EPM: 51.52. Result ID: 6a01b08f47f34ce924e44389.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

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

Recorded evaluation configuration: {"Size":"7B","Frames":"1 fps","Mode":"Online","Training Strategy":"Training-Free"}. This record does not assert independent reproduction.

### LLaVA-Video

EPM: 51.18. Result ID: 6a01b08f47f34ce924e4437c.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

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

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

Recorded evaluation configuration: {"Size":"7B","Frames":"64","Mode":"Offline"}. This record does not assert independent reproduction.

### Qwen2.5-VL-7B + SAVEMem

EPM: 50.84. Result ID: 6a01b08f47f34ce924e4438a.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

Recorded evaluation configuration: {"Size":"7B","Frames":"1 fps","Mode":"Online","Training Strategy":"Training-Free"}. This record does not assert independent reproduction.

### Qwen2.5-VL-3B

EPM: 49.83. Result ID: 6a01b08f47f34ce924e44387.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

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

Recorded evaluation configuration: {"Size":"3B","Frames":"1 fps","Mode":"Online","Training Strategy":"Training-Free"}. This record does not assert independent reproduction.

### Dispider

EPM: 48.48. Result ID: 6a01b08f47f34ce924e44381.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa767f1e209131cea6d744.md>)

[Implementation](<https://github.com/Mark12Ding/Dispider>)

Recorded evaluation configuration: {"Size":"7B","Frames":"1 fps","Mode":"Online","Training Strategy":"Training-Based"}. This record does not assert independent reproduction.

### Qwen2.5-VL-3B + SAVEMem

EPM: 47.81. Result ID: 6a01b08f47f34ce924e44388.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

Recorded evaluation configuration: {"Size":"3B","Frames":"1 fps","Mode":"Online","Training Strategy":"Training-Free"}. This record does not assert independent reproduction.

### Qwen2-VL

EPM: 44.44. Result ID: 6a01b08f47f34ce924e4437d.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75151e209131cea65976.md>)

[Implementation](<https://github.com/QwenLM/Qwen2-VL>)

Recorded evaluation configuration: {"Size":"7B","Frames":"64","Mode":"Offline"}. This record does not assert independent reproduction.

### InternVL2

EPM: 43.1. Result ID: 6a01b08f47f34ce924e4437e.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73081e209131cea59f8b.md>)

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

Recorded evaluation configuration: {"Size":"8B","Frames":"64","Mode":"Offline"}. This record does not assert independent reproduction.

### LongVU

EPM: 43.1. Result ID: 6a01b08f47f34ce924e4437f.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75511e209131cea671f3.md>)

[Implementation](<https://github.com/Vision-CAIR/LongVU>)

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

### Flash-VStream

EPM: 36.36. Result ID: 6a01b08f47f34ce924e44382.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73ea1e209131cea5ede6.md>)

[Implementation](<https://invinciblwyq.github.io/vstream-page>)

Recorded evaluation configuration: {"Size":"7B","Frames":"1 fps","Mode":"Online","Training Strategy":"Training-Based"}. This record does not assert independent reproduction.

### VideoLLM-Online

EPM: 22.22. Result ID: 6a01b08f47f34ce924e44380.

[Reporting paper](<https://www.sota2.com/research/paper/6a0162c2dbc836b87d8233d0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73a71e209131cea5db14.md>)

[Implementation](<https://www.sota2.com/showlab.github.io/videollm-online>)

Recorded evaluation configuration: {"Size":"8B","Frames":"2 fps","Mode":"Online","Training Strategy":"Training-Based"}. This record does not assert independent reproduction.

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