# Multimodal Math Reasoning on MathVerse (avg@8 accuracy)

Source: [Multimodal Math Reasoning on MathVerse (avg@8 accuracy)](<https://www.sota2.com/research/sota/multimodal-math-reasoning-on-mathverse-avg-at-8-accuracy>)

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

Data updated: 2026-04-15T09:12:37.669Z

Evaluation of model performance on the MathVerse dataset for Multimodal Math Reasoning, specifically reporting the average accuracy across 8 attempts (avg@8 accuracy).

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 Math Reasoning |
| Dataset (including any stated split) | MathVerse |
| Metric | Avg@8 Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 10 of 10 |
| 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

- [Avg@8 Accuracy](<https://www.sota2.com/research/sota/multimodal-math-reasoning-on-mathverse-avg-at-8-accuracy.md?metric=Avg%408%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Avg@8 Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| PAPO\_D | 68.58 | Not recorded | 68fa7a761e209131cea8363f |
| PAPO\_G | 68.43 | Not recorded | 68fa7a761e209131cea8363f |
| GRPO | 66.51 | Not recorded | 68fa7a761e209131cea8363f |
| PAPO\_D | 60.51 | Not recorded | 68fa7a761e209131cea8363f |
| PAPO\_G | 57.14 | Not recorded | 68fa7a761e209131cea8363f |
| DAPO | 56.25 | Not recorded | 68fa7a761e209131cea8363f |
| DAPO | 55.64 | Not recorded | 68fa7a761e209131cea8363f |
| GRPO | 55.25 | Not recorded | 68fa7a761e209131cea8363f |
| PAPO\_G | 51.89 | Not recorded | 68fa7a761e209131cea8363f |
| GRPO | 47.98 | Not recorded | 68fa7a761e209131cea8363f |

### PAPO\_D

Avg@8 Accuracy: 68.58. Result ID: 69df54ba47f34ce924c8444e.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

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

### PAPO\_G

Avg@8 Accuracy: 68.43. Result ID: 69df54ba47f34ce924c8444a.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

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

### GRPO

Avg@8 Accuracy: 66.51. Result ID: 69df54ba47f34ce924c84449.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73101e209131cea5a395.md>)

[Implementation](<https://github.com/deepseek-ai/DeepSeek-Math>)

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

### PAPO\_D

Avg@8 Accuracy: 60.51. Result ID: 69df54ba47f34ce924c8444c.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

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

### PAPO\_G

Avg@8 Accuracy: 57.14. Result ID: 69df54ba47f34ce924c84448.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

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

### DAPO

Avg@8 Accuracy: 56.25. Result ID: 69df54ba47f34ce924c8444b.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78a81e209131cea79ac6.md>)

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

### DAPO

Avg@8 Accuracy: 55.64. Result ID: 69df54ba47f34ce924c8444d.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78a81e209131cea79ac6.md>)

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

### GRPO

Avg@8 Accuracy: 55.25. Result ID: 69df54ba47f34ce924c84447.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73101e209131cea5a395.md>)

[Implementation](<https://github.com/deepseek-ai/DeepSeek-Math>)

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

### PAPO\_G

Avg@8 Accuracy: 51.89. Result ID: 69df54ba47f34ce924c84450.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-VL (thinking)","Model Size":"2B"}. This record does not assert independent reproduction.

### GRPO

Avg@8 Accuracy: 47.98. Result ID: 69df54ba47f34ce924c8444f.

[Reporting paper](<https://www.sota2.com/research/paper/68fa7a761e209131cea8363f.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa73101e209131cea5a395.md>)

[Implementation](<https://github.com/deepseek-ai/DeepSeek-Math>)

Recorded evaluation configuration: {"Backbone":"Qwen3-VL (thinking)","Model Size":"2B"}. This record does not assert independent reproduction.

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