# Video Text Generation on VCGBench

Source: [Video Text Generation on VCGBench](<https://www.sota2.com/research/sota/video-text-generation-on-vcgbench>)

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

Data updated: 2026-02-26T05:39:52.785Z

Evaluation of Video Text Generation models using the VCGBench dataset, assessed across dimensions including Correct Information, Detail Orientation, Contextual Understanding, Temporal Understanding, and Consistency.

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 Text Generation |
| Dataset (including any stated split) | VCGBench |
| Metric | Correct Information (CI) |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 14 of 14 |
| 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

- [Correct Information (CI)](<https://www.sota2.com/research/sota/video-text-generation-on-vcgbench.md?metric=Correct%20Information%20(CI)>): Higher is better; unit: SCALAR.

- [Detail Orientation (DO)](<https://www.sota2.com/research/sota/video-text-generation-on-vcgbench.md?metric=Detail%20Orientation%20(DO)>): Higher is better; unit: SCALAR.

- [Contextual Understanding (CU)](<https://www.sota2.com/research/sota/video-text-generation-on-vcgbench.md?metric=Contextual%20Understanding%20(CU)>): Higher is better; unit: SCALAR.

- [Temporal Understanding (TU)](<https://www.sota2.com/research/sota/video-text-generation-on-vcgbench.md?metric=Temporal%20Understanding%20(TU)>): Higher is better; unit: SCALAR.

- [Consistency (CO)](<https://www.sota2.com/research/sota/video-text-generation-on-vcgbench.md?metric=Consistency%20(CO)>): Higher is better; unit: SCALAR.

- [Average Score](<https://www.sota2.com/research/sota/video-text-generation-on-vcgbench.md?metric=Average%20Score>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Correct Information (CI) | Result date | Report paper |
| --- | --- | --- | --- |
| LLaVA-NeXT-Video-DPO | 3.64 | Not recorded | 68fa74c91e209131cea63cb0 |
| LLaVA-NeXT-Video | 3.39 | Not recorded | 68fa74c91e209131cea63cb0 |
| VideoGPT+ | 3.27 | Not recorded | 68fa74c91e209131cea63cb0 |
| Video-LLaMA2 | 3.16 | Not recorded | 68fa74c91e209131cea63cb0 |
| IG-VLM (LLaVA-v1.6) | 3.11 | Not recorded | 68fa74c91e209131cea63cb0 |
| SF-LLaVA-7B | 3.09 | Not recorded | 68fa74c91e209131cea63cb0 |
| LLaVA-NeXT-Image | 3.05 | Not recorded | 68fa74c91e209131cea63cb0 |
| VideoChat2 | 3.02 | Not recorded | 68fa74c91e209131cea63cb0 |
| LLAMA-VID | 2.96 | Not recorded | 68fa74c91e209131cea63cb0 |
| MovieChat | 2.76 | Not recorded | 68fa74c91e209131cea63cb0 |
| Video-ChatGPT | 2.5 | Not recorded | 68fa74c91e209131cea63cb0 |
| Vista-LLaMA | 2.44 | Not recorded | 68fa74c91e209131cea63cb0 |
| VideoChat | 2.23 | Not recorded | 68fa74c91e209131cea63cb0 |
| Video-LLaMA | 1.96 | Not recorded | 68fa74c91e209131cea63cb0 |

### LLaVA-NeXT-Video-DPO

Correct Information (CI): 3.64. Result ID: 69937b953a6c3148491642a9.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"CLIP-L","Training Strategy":"SFT"}. This record does not assert independent reproduction.

### LLaVA-NeXT-Video

Correct Information (CI): 3.39. Result ID: 69937b953a6c3148491642a8.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"CLIP-L","Training Strategy":"SFT"}. This record does not assert independent reproduction.

### VideoGPT+

Correct Information (CI): 3.27. Result ID: 69937b953a6c3148491642a2.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73961e209131cea5d657.md>)

[Implementation](<https://github.com/mbzuai-oryx/VideoGPT-plus>)

Recorded evaluation configuration: {"LLM Size":"3.8B","Vision Encoder":"CLIP-L","Training Strategy":"SFT"}. This record does not assert independent reproduction.

### Video-LLaMA2

Correct Information (CI): 3.16. Result ID: 69937b953a6c3148491642a0.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

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

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

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"CLIP-L","Training Strategy":"SFT"}. This record does not assert independent reproduction.

### IG-VLM (LLaVA-v1.6)

Correct Information (CI): 3.11. Result ID: 69937b953a6c3148491642ab.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72b71e209131cea581e0.md>)

[Implementation](<https://github.com/imagegridworth/IG-VLM>)

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"CLIP-L","Training Strategy":"Training-free"}. This record does not assert independent reproduction.

### SF-LLaVA-7B

Correct Information (CI): 3.09. Result ID: 69937b953a6c3148491642ac.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

[Implementation](<https://huggingface.co/collections/liuhaotian/llava-next-65b9e40155f60fd046a5ccf2>)

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"CLIP-L","Training Strategy":"Training-free"}. This record does not assert independent reproduction.

### LLaVA-NeXT-Image

Correct Information (CI): 3.05. Result ID: 69937b953a6c3148491642aa.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"CLIP-L","Training Strategy":"Training-free"}. This record does not assert independent reproduction.

### VideoChat2

Correct Information (CI): 3.02. Result ID: 69937b953a6c3148491642a5.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73411e209131cea5b4b7.md>)

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"UMT-L","Training Strategy":"SFT"}. This record does not assert independent reproduction.

### LLAMA-VID

Correct Information (CI): 2.96. Result ID: 69937b953a6c3148491642a7.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71461e209131cea50559.md>)

[Implementation](<https://github.com/dvlab-research/LLaMA-VID>)

Recorded evaluation configuration: {"LLM Size":"13B","Vision Encoder":"CLIP-G","Training Strategy":"SFT"}. This record does not assert independent reproduction.

### MovieChat

Correct Information (CI): 2.76. Result ID: 69937b953a6c3148491642a3.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72711e209131cea56563.md>)

[Implementation](<https://rese1f.github.io/MovieChat>)

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"CLIP-G","Training Strategy":"SFT"}. This record does not assert independent reproduction.

### Video-ChatGPT

Correct Information (CI): 2.5. Result ID: 69937b953a6c3148491642a1.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73841e209131cea5ce32.md>)

[Implementation](<https://github.com/mbzuai-oryx/Video-ChatGPT>)

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"CLIP-L","Training Strategy":"SFT"}. This record does not assert independent reproduction.

### Vista-LLaMA

Correct Information (CI): 2.44. Result ID: 69937b953a6c3148491642a6.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa773a1e209131cea71cee.md>)

[Implementation](<https://jinxxian.github.io/Vista-LLaMA>)

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"CLIP-G","Training Strategy":"SFT"}. This record does not assert independent reproduction.

### VideoChat

Correct Information (CI): 2.23. Result ID: 69937b953a6c3148491642a4.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71ba1e209131cea52cc7.md>)

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"CLIP-G","Training Strategy":"SFT"}. This record does not assert independent reproduction.

### Video-LLaMA

Correct Information (CI): 1.96. Result ID: 69937b953a6c31484916429f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa74c91e209131cea63cb0.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa70f11e209131cea4e393.md>)

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

Recorded evaluation configuration: {"LLM Size":"7B","Vision Encoder":"CLIP-G","Training Strategy":"SFT"}. This record does not assert independent reproduction.

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