# Multimodal Reasoning on MMStar (ACC%, S, I)

Source: [Multimodal Reasoning on MMStar (ACC%, S, I)](<https://www.sota2.com/research/sota/multimodal-reasoning-on-mmstar-acc-percent-s-i>)

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

Data updated: 2026-03-17T09:50:42.792Z

Evaluation of model performance on the MMStar dataset for Multimodal Reasoning, reporting Accuracy (ACC%), Score (S), and Index (I) metrics.

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 Reasoning |
| Dataset (including any stated split) | MMStar |
| Metric | Accuracy (ACC%) |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 11 of 11 |
| 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 (ACC%)](<https://www.sota2.com/research/sota/multimodal-reasoning-on-mmstar-acc-percent-s-i.md?metric=Accuracy%20(ACC%25)>): Higher is better; unit: PERCENT.

- [Score (S)](<https://www.sota2.com/research/sota/multimodal-reasoning-on-mmstar-acc-percent-s-i.md?metric=Score%20(S)>): Higher is better; unit: SCALAR.

- [Index (I)](<https://www.sota2.com/research/sota/multimodal-reasoning-on-mmstar-acc-percent-s-i.md?metric=Index%20(I)>): Higher is better; unit: SCALAR.

## Recorded results

| Method | Accuracy (ACC%) | Result date | Report paper |
| --- | --- | --- | --- |
| Qwen3-VL-30B+VRGA | 67.1 | Not recorded | 69b8e0390636e5bf4afbed13 |
| Qwen3-VL-30B | 66.1 | Not recorded | 69b8e0390636e5bf4afbed13 |
| Qwen2-VL-7B | 52.37 | Not recorded | 69b8e0390636e5bf4afbed13 |
| Qwen2-VL-7B+VRGA | 51.07 | Not recorded | 69b8e0390636e5bf4afbed13 |
| Qwen2.5-VL-3B+VRGA | 50.93 | Not recorded | 69b8e0390636e5bf4afbed13 |
| Qwen2-VL-7B+CCOT | 50.06 | Not recorded | 69b8e0390636e5bf4afbed13 |
| Qwen2.5-VL-3B | 49.8 | Not recorded | 69b8e0390636e5bf4afbed13 |
| Qwen2.5-VL-7B+VRGA | 48.73 | Not recorded | 69b8e0390636e5bf4afbed13 |
| Qwen2.5-VL-7B | 48.53 | Not recorded | 69b8e0390636e5bf4afbed13 |
| Qwen2.5-VL-3B+CCOT | 46.27 | Not recorded | 69b8e0390636e5bf4afbed13 |
| Qwen2-VL-7B+ICOT | 42.27 | Not recorded | 69b8e0390636e5bf4afbed13 |

### Qwen3-VL-30B+VRGA

Accuracy (ACC%): 67.1. Result ID: 69b9232147f34ce924a238b3.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-VL-30B","Enhancement":"VRGA"}. This record does not assert independent reproduction.

### Qwen3-VL-30B

Accuracy (ACC%): 66.1. Result ID: 69b9232147f34ce924a238b2.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

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

Recorded evaluation configuration: {"Backbone":"Qwen3-VL-30B"}. This record does not assert independent reproduction.

### Qwen2-VL-7B

Accuracy (ACC%): 52.37. Result ID: 69b9232147f34ce924a238b9.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

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

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

Recorded evaluation configuration: {"Backbone":"Qwen2-VL-7B"}. This record does not assert independent reproduction.

### Qwen2-VL-7B+VRGA

Accuracy (ACC%): 51.07. Result ID: 69b9232147f34ce924a238bc.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL-7B","Enhancement":"VRGA"}. This record does not assert independent reproduction.

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

Accuracy (ACC%): 50.93. Result ID: 69b9232147f34ce924a238b6.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-3B","Enhancement":"VRGA"}. This record does not assert independent reproduction.

### Qwen2-VL-7B+CCOT

Accuracy (ACC%): 50.06. Result ID: 69b9232147f34ce924a238ba.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72be1e209131cea584e9.md>)

[Implementation](<https://github.com/chancharikmitra/CCoT>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL-7B","Enhancement":"CCOT"}. This record does not assert independent reproduction.

### Qwen2.5-VL-3B

Accuracy (ACC%): 49.8. Result ID: 69b9232147f34ce924a238b4.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

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

Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-3B"}. This record does not assert independent reproduction.

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

Accuracy (ACC%): 48.73. Result ID: 69b9232147f34ce924a238b8.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B","Enhancement":"VRGA"}. This record does not assert independent reproduction.

### Qwen2.5-VL-7B

Accuracy (ACC%): 48.53. Result ID: 69b9232147f34ce924a238b7.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

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

Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B"}. This record does not assert independent reproduction.

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

Accuracy (ACC%): 46.27. Result ID: 69b9232147f34ce924a238b5.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72be1e209131cea584e9.md>)

[Implementation](<https://github.com/chancharikmitra/CCoT>)

Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-3B","Enhancement":"CCOT"}. This record does not assert independent reproduction.

### Qwen2-VL-7B+ICOT

Accuracy (ACC%): 42.27. Result ID: 69b9232147f34ce924a238bb.

[Reporting paper](<https://www.sota2.com/research/paper/69b8e0390636e5bf4afbed13.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa777c1e209131cea73816.md>)

Recorded evaluation configuration: {"Backbone":"Qwen2-VL-7B","Enhancement":"ICOT"}. This record does not assert independent reproduction.

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