# Mathematical Reasoning on DAPO + MATH

Source: [Mathematical Reasoning on DAPO + MATH](<https://www.sota2.com/research/sota/mathematical-reasoning-on-dapo-plus-math>)

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

Data updated: 2026-03-27T12:58:33.238Z

Evaluation of model efficiency (training time, rollout time, total time) for mathematical reasoning tasks using the combined DAPO and MATH datasets.

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 | Mathematical Reasoning |
| Dataset (including any stated split) | DAPO + MATH |
| Metric | Training Time (hours) |
| Unit | TIME |
| Direction | Lower is better |
| Results loaded | 12 of 12 |
| 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

- [Training Time (hours)](<https://www.sota2.com/research/sota/mathematical-reasoning-on-dapo-plus-math.md?metric=Training%20Time%20(hours)>): Lower is better; unit: TIME.

- [Other Time (hours)](<https://www.sota2.com/research/sota/mathematical-reasoning-on-dapo-plus-math.md?metric=Other%20Time%20(hours)>): Lower is better; unit: TIME.

- [Rollout Time (hours)](<https://www.sota2.com/research/sota/mathematical-reasoning-on-dapo-plus-math.md?metric=Rollout%20Time%20(hours)>): Lower is better; unit: TIME.

- [Total Time (hours)](<https://www.sota2.com/research/sota/mathematical-reasoning-on-dapo-plus-math.md?metric=Total%20Time%20(hours)>): Lower is better; unit: TIME.

## Recorded results

| Method | Training Time (hours) | Result date | Report paper |
| --- | --- | --- | --- |
| DS | 9.6 | Not recorded | 69c60eff0636e5bf4afc84e3 |
| HIVE | 10.1 | Not recorded | 69c60eff0636e5bf4afc84e3 |
| GRESO | 10.4 | Not recorded | 69c60eff0636e5bf4afc84e3 |
| DS | 10.9 | Not recorded | 69c60eff0636e5bf4afc84e3 |
| HIVE | 11.2 | Not recorded | 69c60eff0636e5bf4afc84e3 |
| GRESO | 11.3 | Not recorded | 69c60eff0636e5bf4afc84e3 |
| DS | 20.6 | Not recorded | 69c60eff0636e5bf4afc84e3 |
| GRESO | 21.8 | Not recorded | 69c60eff0636e5bf4afc84e3 |
| HIVE | 22.2 | Not recorded | 69c60eff0636e5bf4afc84e3 |
| DS | 32.1 | Not recorded | 69c60eff0636e5bf4afc84e3 |
| HIVE | 32.7 | Not recorded | 69c60eff0636e5bf4afc84e3 |
| GRESO | 32.8 | Not recorded | 69c60eff0636e5bf4afc84e3 |

### DS

Training Time (hours): 9.6. Result ID: 69c67e6947f34ce924b07329.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

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

Recorded evaluation configuration: {"Model backbone":"DeepSeek-R1-Distill-Qwen-1.5B","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

### HIVE

Training Time (hours): 10.1. Result ID: 69c67e6947f34ce924b0732b.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

Recorded evaluation configuration: {"Model backbone":"DeepSeek-R1-Distill-Qwen-1.5B","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

### GRESO

Training Time (hours): 10.4. Result ID: 69c67e6947f34ce924b0732a.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79181e209131cea7c1cb.md>)

Recorded evaluation configuration: {"Model backbone":"DeepSeek-R1-Distill-Qwen-1.5B","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

### DS

Training Time (hours): 10.9. Result ID: 69c67e6947f34ce924b07326.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

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

Recorded evaluation configuration: {"Model backbone":"Qwen2.5-Math-1.5B","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

### HIVE

Training Time (hours): 11.2. Result ID: 69c67e6947f34ce924b07328.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

Recorded evaluation configuration: {"Model backbone":"Qwen2.5-Math-1.5B","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

### GRESO

Training Time (hours): 11.3. Result ID: 69c67e6947f34ce924b07327.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79181e209131cea7c1cb.md>)

Recorded evaluation configuration: {"Model backbone":"Qwen2.5-Math-1.5B","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

### DS

Training Time (hours): 20.6. Result ID: 69c67e6947f34ce924b0732c.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

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

Recorded evaluation configuration: {"Model backbone":"Llama3.2-3b-Instruct","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

### GRESO

Training Time (hours): 21.8. Result ID: 69c67e6947f34ce924b0732d.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79181e209131cea7c1cb.md>)

Recorded evaluation configuration: {"Model backbone":"Llama3.2-3b-Instruct","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

### HIVE

Training Time (hours): 22.2. Result ID: 69c67e6947f34ce924b0732e.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

Recorded evaluation configuration: {"Model backbone":"Llama3.2-3b-Instruct","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

### DS

Training Time (hours): 32.1. Result ID: 69c67e6947f34ce924b0732f.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

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

Recorded evaluation configuration: {"Model backbone":"Qwen2.5-Math-7B","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

### HIVE

Training Time (hours): 32.7. Result ID: 69c67e6947f34ce924b07331.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

Recorded evaluation configuration: {"Model backbone":"Qwen2.5-Math-7B","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

### GRESO

Training Time (hours): 32.8. Result ID: 69c67e6947f34ce924b07330.

[Reporting paper](<https://www.sota2.com/research/paper/69c60eff0636e5bf4afc84e3.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa79181e209131cea7c1cb.md>)

Recorded evaluation configuration: {"Model backbone":"Qwen2.5-Math-7B","Training steps":"1000","Hardware (GPUs)":"8x A100"}. This record does not assert independent reproduction.

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