# Multimodal Assistant Evaluation on LLaVA-Bench GPT-4V-aided (full)

Source: [Multimodal Assistant Evaluation on LLaVA-Bench GPT-4V-aided (full)](<https://www.sota2.com/research/sota/multimodal-assistant-evaluation-on-llava-bench-gpt-4v-aided-full>)

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

Data updated: 2026-03-18T08:31:53.116Z

Evaluation of Multimodal Assistant capabilities using the full set of the GPT-4V-aided LLaVA-Bench dataset, focusing on Accuracy and Detailedness scores.

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 Assistant Evaluation |
| Dataset (including any stated split) | LLaVA-Bench GPT-4V-aided (full) |
| Metric | Accuracy |
| Unit | SCALAR |
| Direction | Higher is better |
| Results loaded | 6 of 6 |
| 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/multimodal-assistant-evaluation-on-llava-bench-gpt-4v-aided-full.md?metric=Accuracy>): Higher is better; unit: SCALAR.

- [Detailedness](<https://www.sota2.com/research/sota/multimodal-assistant-evaluation-on-llava-bench-gpt-4v-aided-full.md?metric=Detailedness>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| LTS-FS(Nullu) | 6.96 | Not recorded | 69ba31830636e5bf4afc0754 |
| LTS-FS(Nullu) | 6.59 | Not recorded | 69ba31830636e5bf4afc0754 |
| Nullu | 6.46 | Not recorded | 69ba31830636e5bf4afc0754 |
| Nullu | 6.37 | Not recorded | 69ba31830636e5bf4afc0754 |
| Original | 6.06 | Not recorded | 69ba31830636e5bf4afc0754 |
| Original | 5.74 | Not recorded | 69ba31830636e5bf4afc0754 |

### LTS-FS(Nullu)

Accuracy: 6.96. Result ID: 69ba626347f34ce924a44384.

[Reporting paper](<https://www.sota2.com/research/paper/69ba31830636e5bf4afc0754.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69ba31830636e5bf4afc0754.md>)

Recorded evaluation configuration: {"Model":"LLaVA-1.5"}. This record does not assert independent reproduction.

### LTS-FS(Nullu)

Accuracy: 6.59. Result ID: 69ba626347f34ce924a44387.

[Reporting paper](<https://www.sota2.com/research/paper/69ba31830636e5bf4afc0754.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69ba31830636e5bf4afc0754.md>)

Recorded evaluation configuration: {"Model":"Qwen-vl2.5"}. This record does not assert independent reproduction.

### Nullu

Accuracy: 6.46. Result ID: 69ba626347f34ce924a44383.

[Reporting paper](<https://www.sota2.com/research/paper/69ba31830636e5bf4afc0754.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777c1e209131cea73837.md>)

[Implementation](<https://github.com/Ziwei-Zheng/Nullu>)

Recorded evaluation configuration: {"Model":"LLaVA-1.5"}. This record does not assert independent reproduction.

### Nullu

Accuracy: 6.37. Result ID: 69ba626347f34ce924a44386.

[Reporting paper](<https://www.sota2.com/research/paper/69ba31830636e5bf4afc0754.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777c1e209131cea73837.md>)

[Implementation](<https://github.com/Ziwei-Zheng/Nullu>)

Recorded evaluation configuration: {"Model":"Qwen-vl2.5"}. This record does not assert independent reproduction.

### Original

Accuracy: 6.06. Result ID: 69ba626347f34ce924a44385.

[Reporting paper](<https://www.sota2.com/research/paper/69ba31830636e5bf4afc0754.md>)

Recorded evaluation configuration: {"Model":"Qwen-vl2.5"}. This record does not assert independent reproduction.

### Original

Accuracy: 5.74. Result ID: 69ba626347f34ce924a44382.

[Reporting paper](<https://www.sota2.com/research/paper/69ba31830636e5bf4afc0754.md>)

Recorded evaluation configuration: {"Model":"LLaVA-1.5"}. This record does not assert independent reproduction.

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