# Video Question Answering on EgoSchema (Full)

Source: [Video Question Answering on EgoSchema (Full)](<https://www.sota2.com/research/sota/video-question-answering-on-egoschema-full>)

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

Data updated: 2026-06-30T07:51:47.963Z

Evaluation of model performance on the full set of the EgoSchema dataset for Video Question Answering.

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) | EgoSchema (Full) |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 256 of 256 |
| 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/video-question-answering-on-egoschema-full.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| TASKER | 77.3 | Not recorded | 6a434d7cba73165dbb743955 |
| VideoTree | 76.7 | Not recorded | 6a434d7cba73165dbb743955 |
| Human Eval | 75 | Not recorded | 68fa75901e209131cea68a54 |
| HCQA | 75 | Not recorded | 68fa785c1e209131cea77f75 |
| VideoAgent | 74.6 | Not recorded | 6a434d7cba73165dbb743955 |
| GPT-4o | 72.2 | Not recorded | 6a0fe301f984739ed6739ccc |
| LinVT-Qwen2-VL | 69.5 | Not recorded | 68fa785c1e209131cea77f75 |
| ProVCA | 69.3 | Not recorded | 69d33de37d0172140ec73565 |
| VideoMultiAgents | 68 | Not recorded | 68fa785c1e209131cea77f75 |
| LongVU | 67.6 | Not recorded | 68fa785c1e209131cea77f75 |
| LongVU | 67.6 | Not recorded | 68fa777a1e209131cea73736 |
| Qwen3-VL-4B + MuKV | 67 | Not recorded | 6a0fe301f984739ed6739ccc |
| Qwen3-VL-2B + MuKV | 66.3 | Not recorded | 6a0fe301f984739ed6739ccc |
| Qwen3-VL-4B | 65.8 | Not recorded | 6a0fe301f984739ed6739ccc |
| Gemini 1.5 Flash | 65.7 | Not recorded | 68fa76b61e209131cea6ea89 |
| Gemini 1.5 Flash | 65.7 | Not recorded | 68fa76b61e209131cea6ea89 |
| VideoINSTA | 65 | Not recorded | 68fa76b61e209131cea6ea89 |
| VideoINSTA | 65 | Not recorded | 68fa76b61e209131cea6ea89 |
| Qwen3-VL-2B | 65 | Not recorded | 6a0fe301f984739ed6739ccc |
| Lifelong Memory | 64.7 | Not recorded | 68fa785c1e209131cea77f75 |
| Qwen2.5-VL-3B | 64.4 | Not recorded | 6a0fe301f984739ed6739ccc |
| VideoLLaMA2 | 63.9 | Not recorded | 68fa785c1e209131cea77f75 |
| VideoLLaMA 2 | 63.9 | Not recorded | 68fa777a1e209131cea73736 |
| AKEYS | 63.6 | Not recorded | 68fa77801e209131cea739ed |
| TASKER | 63.6 | Not recorded | 6a434d7cba73165dbb743955 |
| InternVL-3 + A.I.R. | 63.3 | Not recorded | 68fa7be01e209131cea8af78 |
| LLaVA-OV-7B + MuKV | 63.3 | Not recorded | 6a0fe301f984739ed6739ccc |
| Gemini-1.5-Pro | 63.2 | Not recorded | 68fa75901e209131cea68a54 |
| AKEYS | 63.1 | Not recorded | 68fa77801e209131cea739ed |
| TASKER | 63.1 | Not recorded | 6a434d7cba73165dbb743955 |
| LLaVA-OV-7B + LiveVLM | 63 | Not recorded | 6a0fe301f984739ed6739ccc |
| LLaVA-OV-7B + StreamMem | 63 | Not recorded | 6a0fe301f984739ed6739ccc |
| Qwen2.5-VL-3B + MuKV | 63 | Not recorded | 6a0fe301f984739ed6739ccc |
| InternVL-3 | 62.5 | Not recorded | 68fa7be01e209131cea8af78 |
| LifelongMemory | 62.4 | Not recorded | 68fa75901e209131cea68a54 |
| Qwen2.5-VL-3B + StreamMem | 62.2 | Not recorded | 6a0fe301f984739ed6739ccc |
| LifelongMemory | 62.1 | Not recorded | 68fa764b1e209131cea6c30b |
| LifelongMemory | 62.1 | Not recorded | 68fa77981e209131cea7438f |
| LLaVA-OV-72B | 62 | Not recorded | 68fa777a1e209131cea73736 |
| Qwen2.5-VL-3B + InfiniPot-V | 61.8 | Not recorded | 6a0fe301f984739ed6739ccc |
| Tarsier-34B | 61.7 | Not recorded | 68fa74e41e209131cea64782 |
| Tarsier | 61.7 | Not recorded | 68fa764b1e209131cea6c30b |
| Tarsier | 61.7 | Not recorded | 68fa77981e209131cea7438f |
| Tarsier | 61.7 | Not recorded | 68fa785c1e209131cea77f75 |
| Tarsier-34B | 61.7 | Not recorded | 68fa777a1e209131cea73736 |
| InternVL-3 + MDP3 | 61.6 | Not recorded | 68fa7be01e209131cea8af78 |
| LLaVA-OneVision + A.I.R. | 61.4 | Not recorded | 68fa7be01e209131cea8af78 |
| VideoTree | 61.1 | Not recorded | 68fa74e41e209131cea64782 |
| LVNet | 61.1 | Not recorded | 68fa764b1e209131cea6c30b |
| VideoTree | 61.1 | Not recorded | 68fa764b1e209131cea6c30b |
| VideoTree | 61.1 | Not recorded | 68fa76b61e209131cea6ea89 |
| LVNet | 61.1 | Not recorded | 68fa76b61e209131cea6ea89 |
| VideoTree | 61.1 | Not recorded | 68fa76b61e209131cea6ea89 |
| LVNet | 61.1 | Not recorded | 68fa76b61e209131cea6ea89 |
| LVNet | 61.1 | Not recorded | 68fa77801e209131cea739ed |
| VideoTree | 61.1 | Not recorded | 68fa77801e209131cea739ed |
| VideoTree | 61.1 | Not recorded | 68fa77981e209131cea7438f |
| LVNet | 61.1 | Not recorded | 68fa77981e209131cea7438f |
| VideoTree | 61.1 | Not recorded | 68fa785c1e209131cea77f75 |
| LVNet | 61.1 | Not recorded | 68fa785c1e209131cea77f75 |
| VIDEOTREE | 61.1 | Not recorded | 68fa777a1e209131cea73736 |
| VideoTree | 61.1 | Not recorded | 68fa7be01e209131cea8af78 |
| VideoTree | 61.1 | Not recorded | 69d33de37d0172140ec73565 |
| LVNet | 61.1 | Not recorded | 69d33de37d0172140ec73565 |
| LVNet | 61.1 | Not recorded | 6a434d7cba73165dbb743955 |
| VideoTree | 61.1 | Not recorded | 6a434d7cba73165dbb743955 |
| DrVideo | 61 | Not recorded | 68fa7be01e209131cea8af78 |
| DrVideo | 61 | Not recorded | 69d33de37d0172140ec73565 |
| LLaVA-OneVision + BOLT | 60.7 | Not recorded | 68fa7be01e209131cea8af78 |
| LLaVA-OV-7B + ReKV | 60.7 | Not recorded | 6a0fe301f984739ed6739ccc |
| LLaVA-OneVision 32 frames + BOLT | 60.66 | Not recorded | 68fa779c1e209131cea7457f |
| LLaVA-OneVision 32 frames | 60.36 | Not recorded | 68fa779c1e209131cea7457f |
| LLaVA-OneVision + MDP3 | 60.3 | Not recorded | 68fa7be01e209131cea8af78 |
| VideoAgent | 60.2 | Not recorded | 68fa75901e209131cea68a54 |
| InternVideo2 | 60.2 | Not recorded | 68fa764b1e209131cea6c30b |
| VideoAgent \[6\] | 60.2 | Not recorded | 68fa77801e209131cea739ed |
| InternVideo2 | 60.2 | Not recorded | 68fa77981e209131cea7438f |
| VideoChat2 | 60.2 | Not recorded | 68fa777a1e209131cea73736 |
| LLaVA-OneVision | 60.2 | Not recorded | 68fa7be01e209131cea8af78 |
| VideoAgent | 60.2 | Not recorded | 69d33de37d0172140ec73565 |

### TASKER

Accuracy: 77.3. Result ID: 6a437579fbc373592c93ee07.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7cba73165dbb743955.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a434d7cba73165dbb743955.md>)

[Implementation](<https://github.com/VG-GUI-TASKER/VG-GUI-TASKER>)

Recorded evaluation configuration: {"(M)LLM":"Qwen3-VL"}. This record does not assert independent reproduction.

### VideoTree

Accuracy: 76.7. Result ID: 6a437579fbc373592c93ee06.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7cba73165dbb743955.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Implementation](<https://videotree2024.github.io/>)

Recorded evaluation configuration: {"(M)LLM":"Qwen3-VL"}. This record does not assert independent reproduction.

### Human Eval

Accuracy: 75. Result ID: 6993edec3a6c314849198c6e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75901e209131cea68a54.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70561e209131cea4a59b.md>)

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

### HCQA

Accuracy: 75. Result ID: 6995877e3badfaffe6807527.

[Reporting paper](<https://www.sota2.com/research/paper/68fa785c1e209131cea77f75.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa756a1e209131cea67bc8.md>)

[Implementation](<https://github.com/Hyu-Zhang/HCQA>)

Recorded evaluation configuration: {"mode":"Zero-shot"}. This record does not assert independent reproduction.

### VideoAgent

Accuracy: 74.6. Result ID: 6a437579fbc373592c93ee05.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7cba73165dbb743955.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72971e209131cea5761e.md>)

[Implementation](<https://wxh1996.github.io/VideoAgent-Website/>)

Recorded evaluation configuration: {"(M)LLM":"Qwen3-VL"}. This record does not assert independent reproduction.

### GPT-4o

Accuracy: 72.2. Result ID: 6a1015adfbc373592c091f9a.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

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

Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction.

### LinVT-Qwen2-VL

Accuracy: 69.5. Result ID: 6995877e3badfaffe6807526.

[Reporting paper](<https://www.sota2.com/research/paper/68fa785c1e209131cea77f75.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa762f1e209131cea6bbab.md>)

[Implementation](<https://github.com/gls0425/LinVT>)

Recorded evaluation configuration: {"mode":"Zero-shot"}. This record does not assert independent reproduction.

### ProVCA

Accuracy: 69.3. Result ID: 69d36f8547f34ce924bb47a7.

[Reporting paper](<https://www.sota2.com/research/paper/69d33de37d0172140ec73565.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69d33de37d0172140ec73565.md>)

Recorded evaluation configuration: {"LLM/MLLM":"GPT-4o"}. This record does not assert independent reproduction.

### VideoMultiAgents

Accuracy: 68. Result ID: 6995877e3badfaffe6807528.

[Reporting paper](<https://www.sota2.com/research/paper/68fa785c1e209131cea77f75.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa785c1e209131cea77f75.md>)

[Implementation](<https://github.com/PanasonicConnect/VideoMultiAgents>)

Recorded evaluation configuration: {"mode":"Zero-shot"}. This record does not assert independent reproduction.

### LongVU

Accuracy: 67.6. Result ID: 6995877e3badfaffe6807525.

[Reporting paper](<https://www.sota2.com/research/paper/68fa785c1e209131cea77f75.md>)

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

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

Recorded evaluation configuration: {"mode":"Zero-shot"}. This record does not assert independent reproduction.

### LongVU

Accuracy: 67.6. Result ID: 699aafd25698665ec43d2884.

[Reporting paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

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

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

Recorded evaluation configuration: {"Training Status":"Trained"}. This record does not assert independent reproduction.

### Qwen3-VL-4B + MuKV

Accuracy: 67. Result ID: 6a1015adfbc373592c091fab.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

Recorded evaluation configuration: {"Frames/FPS":"0.5fps","# Mem. Tok":"5.9K"}. This record does not assert independent reproduction.

### Qwen3-VL-2B + MuKV

Accuracy: 66.3. Result ID: 6a1015adfbc373592c091fad.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

Recorded evaluation configuration: {"Frames/FPS":"0.5fps","# Mem. Tok":"5.9K"}. This record does not assert independent reproduction.

### Qwen3-VL-4B

Accuracy: 65.8. Result ID: 6a1015adfbc373592c091faa.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa788e1e209131cea79116.md>)

Recorded evaluation configuration: {"Frames/FPS":"768"}. This record does not assert independent reproduction.

### Gemini 1.5 Flash

Accuracy: 65.7. Result ID: 69945f423a6c3148491cd18e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76b61e209131cea6ea89.md>)

Recorded evaluation configuration: {"Architecture Type":"End-to-End","zero-shot":true}. This record does not assert independent reproduction.

### Gemini 1.5 Flash

Accuracy: 65.7. Result ID: 69945f423a6c3148491cd19f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76b61e209131cea6ea89.md>)

Recorded evaluation configuration: {"Paradigm":"End-to-End","Zero-shot":true}. This record does not assert independent reproduction.

### VideoINSTA

Accuracy: 65. Result ID: 69945f423a6c3148491cd192.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76b61e209131cea6ea89.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75131e209131cea658b6.md>)

Recorded evaluation configuration: {"Architecture Type":"Bottom-up","zero-shot":true}. This record does not assert independent reproduction.

### VideoINSTA

Accuracy: 65. Result ID: 69945f433a6c3148491cd1a3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76b61e209131cea6ea89.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75131e209131cea658b6.md>)

Recorded evaluation configuration: {"Paradigm":"Bottom-up","Zero-shot":true}. This record does not assert independent reproduction.

### Qwen3-VL-2B

Accuracy: 65. Result ID: 6a1015adfbc373592c091fac.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa788e1e209131cea79116.md>)

Recorded evaluation configuration: {"Frames/FPS":"768"}. This record does not assert independent reproduction.

### Lifelong Memory

Accuracy: 64.7. Result ID: 6995877e3badfaffe6807524.

[Reporting paper](<https://www.sota2.com/research/paper/68fa785c1e209131cea77f75.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa758c1e209131cea6887b.md>)

Recorded evaluation configuration: {"mode":"Zero-shot"}. This record does not assert independent reproduction.

### Qwen2.5-VL-3B

Accuracy: 64.4. Result ID: 6a1015adfbc373592c091fa6.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

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

Recorded evaluation configuration: {"Frames/FPS":"768"}. This record does not assert independent reproduction.

### VideoLLaMA2

Accuracy: 63.9. Result ID: 6995877e3badfaffe6807523.

[Reporting paper](<https://www.sota2.com/research/paper/68fa785c1e209131cea77f75.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75761e209131cea681dd.md>)

[Implementation](<https://github.com/DAMO-NLP-SG/VideoLLaMA2>)

Recorded evaluation configuration: {"mode":"Zero-shot"}. This record does not assert independent reproduction.

### VideoLLaMA 2

Accuracy: 63.9. Result ID: 699aafd25698665ec43d2882.

[Reporting paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75761e209131cea681dd.md>)

[Implementation](<https://github.com/DAMO-NLP-SG/VideoLLaMA2>)

Recorded evaluation configuration: {"Training Status":"Trained"}. This record does not assert independent reproduction.

### AKEYS

Accuracy: 63.6. Result ID: 69953aa03badfaffe622667e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77801e209131cea739ed.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77801e209131cea739ed.md>)

[Implementation](<https://github.com/fansunqi/AKeyS>)

Recorded evaluation configuration: {"(M)LLM":"GPT-4o"}. This record does not assert independent reproduction.

### TASKER

Accuracy: 63.6. Result ID: 6a437579fbc373592c93ee04.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7cba73165dbb743955.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a434d7cba73165dbb743955.md>)

[Implementation](<https://github.com/VG-GUI-TASKER/VG-GUI-TASKER>)

Recorded evaluation configuration: {"(M)LLM":"GPT-4o"}. This record does not assert independent reproduction.

### InternVL-3 + A.I.R.

Accuracy: 63.3. Result ID: 69ca18fb47f34ce924b208e3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af78.md>)

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

### LLaVA-OV-7B + MuKV

Accuracy: 63.3. Result ID: 6a1015adfbc373592c091fa2.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

Recorded evaluation configuration: {"Frames/FPS":"0.5fps","# Mem. Tok":"5.9K"}. This record does not assert independent reproduction.

### Gemini-1.5-Pro

Accuracy: 63.2. Result ID: 6993edec3a6c314849198c6f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75901e209131cea68a54.md>)

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

Recorded evaluation configuration: {"Core LLMs":"Gemini-1.5-Pro","Zero-shot":"true"}. This record does not assert independent reproduction.

### AKEYS

Accuracy: 63.1. Result ID: 69953aa03badfaffe622667d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77801e209131cea739ed.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77801e209131cea739ed.md>)

[Implementation](<https://github.com/fansunqi/AKeyS>)

Recorded evaluation configuration: {"(M)LLM":"GPT-4"}. This record does not assert independent reproduction.

### TASKER

Accuracy: 63.1. Result ID: 6a437579fbc373592c93ee03.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7cba73165dbb743955.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a434d7cba73165dbb743955.md>)

[Implementation](<https://github.com/VG-GUI-TASKER/VG-GUI-TASKER>)

Recorded evaluation configuration: {"(M)LLM":"GPT-4"}. This record does not assert independent reproduction.

### LLaVA-OV-7B + LiveVLM

Accuracy: 63. Result ID: 6a1015adfbc373592c091fa0.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78a61e209131cea799bb.md>)

Recorded evaluation configuration: {"Frames/FPS":"0.5/0.2 fps"}. This record does not assert independent reproduction.

### LLaVA-OV-7B + StreamMem

Accuracy: 63. Result ID: 6a1015adfbc373592c091fa1.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7ad51e209131cea85682.md>)

Recorded evaluation configuration: {"Frames/FPS":"0.5/0.2 fps","# Mem. Tok":"6K"}. This record does not assert independent reproduction.

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

Accuracy: 63. Result ID: 6a1015adfbc373592c091fa9.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

Recorded evaluation configuration: {"Frames/FPS":"0.5 fps","# Mem. Tok":"5.9K"}. This record does not assert independent reproduction.

### InternVL-3

Accuracy: 62.5. Result ID: 69ca18fb47f34ce924b208e1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78241e209131cea76fa5.md>)

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

### LifelongMemory

Accuracy: 62.4. Result ID: 6993edec3a6c314849198c7a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75901e209131cea68a54.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa758c1e209131cea6887b.md>)

Recorded evaluation configuration: {"category":"LLM-based video agents","Core VLMs":"LaViLa","Core LLMs":"GPT-4","in-domain training (Ego4D)":"true","Zero-shot":"true"}. This record does not assert independent reproduction.

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

Accuracy: 62.2. Result ID: 6a1015adfbc373592c091fa8.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7ad51e209131cea85682.md>)

Recorded evaluation configuration: {"Frames/FPS":"4.0/0.5 fps","# Mem. Tok":"6K"}. This record does not assert independent reproduction.

### LifelongMemory

Accuracy: 62.1. Result ID: 699418813a6c3148491b05c3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa764b1e209131cea6c30b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa758c1e209131cea6887b.md>)

Recorded evaluation configuration: {"Evaluation Protocol":"zero-shot (with proprietary LLMs)","Video Pretrain":false,"Params":"1.8T"}. This record does not assert independent reproduction.

### LifelongMemory

Accuracy: 62.1. Result ID: 6995496d3badfaffe638741b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa758c1e209131cea6887b.md>)

Recorded evaluation configuration: {"Video-level training free":true,"Number of captions":"90"}. This record does not assert independent reproduction.

### LLaVA-OV-72B

Accuracy: 62. Result ID: 699aafd25698665ec43d2883.

[Reporting paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

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

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

Recorded evaluation configuration: {"Training Status":"Trained"}. This record does not assert independent reproduction.

### Qwen2.5-VL-3B + InfiniPot-V

Accuracy: 61.8. Result ID: 6a1015adfbc373592c091fa7.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa798c1e209131cea7e234.md>)

Recorded evaluation configuration: {"Frames/FPS":"768","# Mem. Tok":"6K"}. This record does not assert independent reproduction.

### Tarsier-34B

Accuracy: 61.7. Result ID: 69937e533a6c3148491656ab.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74e41e209131cea64782.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74e41e209131cea64782.md>)

[Implementation](<https://github.com/bytedance/tarsier>)

Recorded evaluation configuration: {"Mode":"Zero-shot"}. This record does not assert independent reproduction.

### Tarsier

Accuracy: 61.7. Result ID: 699418813a6c3148491b05cd.

[Reporting paper](<https://www.sota2.com/research/paper/68fa764b1e209131cea6c30b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74e41e209131cea64782.md>)

[Implementation](<https://github.com/bytedance/tarsier>)

Recorded evaluation configuration: {"Evaluation Protocol":"zero-shot (with open-source LLMs)","Video Pretrain":true,"Params":"34B"}. This record does not assert independent reproduction.

### Tarsier

Accuracy: 61.7. Result ID: 6995496d3badfaffe6387410.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74e41e209131cea64782.md>)

[Implementation](<https://github.com/bytedance/tarsier>)

Recorded evaluation configuration: {"Video-level training free":false}. This record does not assert independent reproduction.

### Tarsier

Accuracy: 61.7. Result ID: 6995877e3badfaffe6807522.

[Reporting paper](<https://www.sota2.com/research/paper/68fa785c1e209131cea77f75.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74e41e209131cea64782.md>)

[Implementation](<https://github.com/bytedance/tarsier>)

Recorded evaluation configuration: {"mode":"Zero-shot"}. This record does not assert independent reproduction.

### Tarsier-34B

Accuracy: 61.7. Result ID: 699aafd25698665ec43d2880.

[Reporting paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74e41e209131cea64782.md>)

[Implementation](<https://github.com/bytedance/tarsier>)

Recorded evaluation configuration: {"Training Status":"Trained"}. This record does not assert independent reproduction.

### InternVL-3 + MDP3

Accuracy: 61.6. Result ID: 69ca18fb47f34ce924b208e2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa767f1e209131cea6d6ec.md>)

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

### LLaVA-OneVision + A.I.R.

Accuracy: 61.4. Result ID: 69ca18fb47f34ce924b208e7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af78.md>)

Recorded evaluation configuration: {"VLM Size":"7B","#Frames":"≤ 32"}. This record does not assert independent reproduction.

### VideoTree

Accuracy: 61.1. Result ID: 69937e533a6c3148491656a9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74e41e209131cea64782.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Implementation](<https://videotree2024.github.io/>)

Recorded evaluation configuration: {"Mode":"Zero-shot"}. This record does not assert independent reproduction.

### LVNet

Accuracy: 61.1. Result ID: 699418813a6c3148491b05bf.

[Reporting paper](<https://www.sota2.com/research/paper/68fa764b1e209131cea6c30b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Implementation](<https://github.com/jongwoopark7978/LVNet>)

Recorded evaluation configuration: {"Evaluation Protocol":"zero-shot (with proprietary LLMs)","Video Pretrain":false,"Params":"\<1.8T"}. This record does not assert independent reproduction.

### VideoTree

Accuracy: 61.1. Result ID: 699418813a6c3148491b05c2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa764b1e209131cea6c30b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Implementation](<https://videotree2024.github.io/>)

Recorded evaluation configuration: {"Evaluation Protocol":"zero-shot (with proprietary LLMs)","Video Pretrain":false,"Params":"1.8T"}. This record does not assert independent reproduction.

### VideoTree

Accuracy: 61.1. Result ID: 69945f423a6c3148491cd190.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76b61e209131cea6ea89.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Implementation](<https://videotree2024.github.io/>)

Recorded evaluation configuration: {"Architecture Type":"Bottom-up","zero-shot":true}. This record does not assert independent reproduction.

### LVNet

Accuracy: 61.1. Result ID: 69945f423a6c3148491cd191.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76b61e209131cea6ea89.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Implementation](<https://github.com/jongwoopark7978/LVNet>)

Recorded evaluation configuration: {"Architecture Type":"Bottom-up","zero-shot":true}. This record does not assert independent reproduction.

### VideoTree

Accuracy: 61.1. Result ID: 69945f433a6c3148491cd1a1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76b61e209131cea6ea89.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Implementation](<https://videotree2024.github.io/>)

Recorded evaluation configuration: {"Paradigm":"Bottom-up","Zero-shot":true}. This record does not assert independent reproduction.

### LVNet

Accuracy: 61.1. Result ID: 69945f433a6c3148491cd1a2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa76b61e209131cea6ea89.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Implementation](<https://github.com/jongwoopark7978/LVNet>)

Recorded evaluation configuration: {"Paradigm":"Bottom-up","Zero-shot":true}. This record does not assert independent reproduction.

### LVNet

Accuracy: 61.1. Result ID: 69953aa03badfaffe6226676.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77801e209131cea739ed.md>)

Recorded evaluation configuration: {"(M)LLM":"GPT-4o"}. This record does not assert independent reproduction.

### VideoTree

Accuracy: 61.1. Result ID: 69953aa03badfaffe622667c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77801e209131cea739ed.md>)

Recorded evaluation configuration: {"(M)LLM":"GPT-4"}. This record does not assert independent reproduction.

### VideoTree

Accuracy: 61.1. Result ID: 6995496d3badfaffe6387419.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Implementation](<https://videotree2024.github.io/>)

Recorded evaluation configuration: {"Video-level training free":true,"Number of captions":"62.4"}. This record does not assert independent reproduction.

### LVNet

Accuracy: 61.1. Result ID: 6995496d3badfaffe638741f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Implementation](<https://github.com/jongwoopark7978/LVNet>)

Recorded evaluation configuration: {"Video-level training free":true,"Number of captions":"12"}. This record does not assert independent reproduction.

### VideoTree

Accuracy: 61.1. Result ID: 6995877e3badfaffe6807520.

[Reporting paper](<https://www.sota2.com/research/paper/68fa785c1e209131cea77f75.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Implementation](<https://videotree2024.github.io/>)

Recorded evaluation configuration: {"mode":"Zero-shot"}. This record does not assert independent reproduction.

### LVNet

Accuracy: 61.1. Result ID: 6995877e3badfaffe6807521.

[Reporting paper](<https://www.sota2.com/research/paper/68fa785c1e209131cea77f75.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Implementation](<https://github.com/jongwoopark7978/LVNet>)

Recorded evaluation configuration: {"mode":"Zero-shot"}. This record does not assert independent reproduction.

### VIDEOTREE

Accuracy: 61.1. Result ID: 699aafd25698665ec43d2885.

[Reporting paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Implementation](<https://videotree2024.github.io/>)

Recorded evaluation configuration: {"Training Status":"Training-free"}. This record does not assert independent reproduction.

### VideoTree

Accuracy: 61.1. Result ID: 69ca18fb47f34ce924b208d8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Implementation](<https://videotree2024.github.io/>)

Recorded evaluation configuration: {"VLM Size":"GPT4"}. This record does not assert independent reproduction.

### VideoTree

Accuracy: 61.1. Result ID: 69d36f8547f34ce924bb47a2.

[Reporting paper](<https://www.sota2.com/research/paper/69d33de37d0172140ec73565.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Implementation](<https://videotree2024.github.io/>)

Recorded evaluation configuration: {"LLM/MLLM":"GPT-4"}. This record does not assert independent reproduction.

### LVNet

Accuracy: 61.1. Result ID: 69d36f8547f34ce924bb47a6.

[Reporting paper](<https://www.sota2.com/research/paper/69d33de37d0172140ec73565.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Implementation](<https://github.com/jongwoopark7978/LVNet>)

Recorded evaluation configuration: {"LLM/MLLM":"GPT-4o"}. This record does not assert independent reproduction.

### LVNet

Accuracy: 61.1. Result ID: 6a437579fbc373592c93edfc.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7cba73165dbb743955.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Implementation](<https://github.com/jongwoopark7978/LVNet>)

Recorded evaluation configuration: {"(M)LLM":"GPT-4o"}. This record does not assert independent reproduction.

### VideoTree

Accuracy: 61.1. Result ID: 6a437579fbc373592c93ee02.

[Reporting paper](<https://www.sota2.com/research/paper/6a434d7cba73165dbb743955.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Implementation](<https://videotree2024.github.io/>)

Recorded evaluation configuration: {"(M)LLM":"GPT-4"}. This record does not assert independent reproduction.

### DrVideo

Accuracy: 61. Result ID: 69ca18fb47f34ce924b208da.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d11e209131cea6a009.md>)

Recorded evaluation configuration: {"VLM Size":"GPT4"}. This record does not assert independent reproduction.

### DrVideo

Accuracy: 61. Result ID: 69d36f8547f34ce924bb47a4.

[Reporting paper](<https://www.sota2.com/research/paper/69d33de37d0172140ec73565.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75d11e209131cea6a009.md>)

Recorded evaluation configuration: {"LLM/MLLM":"GPT-4"}. This record does not assert independent reproduction.

### LLaVA-OneVision + BOLT

Accuracy: 60.7. Result ID: 69ca18fb47f34ce924b208e6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa779c1e209131cea7457f.md>)

[Implementation](<https://github.com/sming256/BOLT>)

Recorded evaluation configuration: {"VLM Size":"7B","#Frames":"32"}. This record does not assert independent reproduction.

### LLaVA-OV-7B + ReKV

Accuracy: 60.7. Result ID: 6a1015adfbc373592c091f9f.

[Reporting paper](<https://www.sota2.com/research/paper/6a0fe301f984739ed6739ccc.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77431e209131cea72135.md>)

Recorded evaluation configuration: {"Frames/FPS":"0.5 fps","# Mem. Tok":"5.9K"}. This record does not assert independent reproduction.

### LLaVA-OneVision 32 frames + BOLT

Accuracy: 60.66. Result ID: 699550723badfaffe6401fbc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa779c1e209131cea7457f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa779c1e209131cea7457f.md>)

[Implementation](<https://github.com/sming256/BOLT>)

Recorded evaluation configuration: {"LLM Size":"7B","Training Free":"true","Frames":32}. This record does not assert independent reproduction.

### LLaVA-OneVision 32 frames

Accuracy: 60.36. Result ID: 699550723badfaffe6401fbb.

[Reporting paper](<https://www.sota2.com/research/paper/68fa779c1e209131cea7457f.md>)

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

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

Recorded evaluation configuration: {"LLM Size":"7B","Training Free":"false","Frames":32}. This record does not assert independent reproduction.

### LLaVA-OneVision + MDP3

Accuracy: 60.3. Result ID: 69ca18fb47f34ce924b208e5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af78.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa767f1e209131cea6d6ec.md>)

Recorded evaluation configuration: {"VLM Size":"7B","#Frames":"32"}. This record does not assert independent reproduction.

### VideoAgent

Accuracy: 60.2. Result ID: 6993edec3a6c314849198c7b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa75901e209131cea68a54.md>)

Recorded evaluation configuration: {"category":"LLM-based video agents","Core VLMs":"Video-LLaVA \[32\]","Core LLMs":"GPT-4","citation":"\[13\]","Zero-shot":"true"}. This record does not assert independent reproduction.

### InternVideo2

Accuracy: 60.2. Result ID: 699418813a6c3148491b05cc.

[Reporting paper](<https://www.sota2.com/research/paper/68fa764b1e209131cea6c30b.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74871e209131cea623f3.md>)

[Implementation](<https://github.com/OpenGVLab/InternVideo2>)

Recorded evaluation configuration: {"Evaluation Protocol":"zero-shot (with open-source LLMs)","Video Pretrain":true,"Params":"13B"}. This record does not assert independent reproduction.

### VideoAgent \[6\]

Accuracy: 60.2. Result ID: 69953aa03badfaffe622667a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77801e209131cea739ed.md>)

Recorded evaluation configuration: {"(M)LLM":"GPT-4"}. This record does not assert independent reproduction.

### InternVideo2

Accuracy: 60.2. Result ID: 6995496d3badfaffe638740f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa77981e209131cea7438f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74871e209131cea623f3.md>)

[Implementation](<https://github.com/OpenGVLab/InternVideo2>)

Recorded evaluation configuration: {"Video-level training free":false}. This record does not assert independent reproduction.

### VideoChat2

Accuracy: 60.2. Result ID: 699aafd25698665ec43d2881.

[Reporting paper](<https://www.sota2.com/research/paper/68fa777a1e209131cea73736.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74871e209131cea623f3.md>)

[Implementation](<https://github.com/OpenGVLab/InternVideo2>)

Recorded evaluation configuration: {"Training Status":"Trained"}. This record does not assert independent reproduction.

### LLaVA-OneVision

Accuracy: 60.2. Result ID: 69ca18fb47f34ce924b208e4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7be01e209131cea8af78.md>)

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

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

Recorded evaluation configuration: {"VLM Size":"7B","#Frames":"32"}. This record does not assert independent reproduction.

### VideoAgent

Accuracy: 60.2. Result ID: 69d36f8547f34ce924bb47a3.

[Reporting paper](<https://www.sota2.com/research/paper/69d33de37d0172140ec73565.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72971e209131cea5761e.md>)

[Implementation](<https://wxh1996.github.io/VideoAgent-Website/>)

Recorded evaluation configuration: {"LLM/MLLM":"GPT-4","Reference":"\[28\]"}. This record does not assert independent reproduction.

Page 1 of 4. This is an index page, not the full collection.

- [Next index page](<https://www.sota2.com/research/sota/video-question-answering-on-egoschema-full.md?metric=Accuracy&page=2>): Follow this link to discover more pages.
