# Mathematical Reasoning on MATH 500 Hard

Source: [Mathematical Reasoning on MATH 500 Hard](<https://www.sota2.com/research/sota/mathematical-reasoning-on-math-500-hard>)

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

Data updated: 2026-02-26T05:32:32.888Z

Evaluation of model performance on the 'Hard' subset of the MATH 500 dataset for mathematical reasoning, measured by 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 | Mathematical Reasoning |
| Dataset (including any stated split) | MATH 500 Hard |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 15 of 15 |
| 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/mathematical-reasoning-on-math-500-hard.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| ReAct | 80.8 | Not recorded | 693266f2d6b0dd33429341b2 |
| PPO | 73.4 | Not recorded | 693266f2d6b0dd33429341b2 |
| NLAC | 72.7 | Not recorded | 693266f2d6b0dd33429341b2 |
| GRPO | 71.8 | Not recorded | 693266f2d6b0dd33429341b2 |
| NLRL | 71.5 | Not recorded | 693266f2d6b0dd33429341b2 |
| NLQL | 71.2 | Not recorded | 693266f2d6b0dd33429341b2 |
| Self-Distillation | 70.1 | Not recorded | 693266f2d6b0dd33429341b2 |
| RFT | 69.4 | Not recorded | 693266f2d6b0dd33429341b2 |
| Self-Distillation | 57.1 | Not recorded | 693266f2d6b0dd33429341b2 |
| NLQL | 56.4 | Not recorded | 693266f2d6b0dd33429341b2 |
| NLAC | 56.2 | Not recorded | 693266f2d6b0dd33429341b2 |
| NLRL | 53.4 | Not recorded | 693266f2d6b0dd33429341b2 |
| GRPO | 52.5 | Not recorded | 693266f2d6b0dd33429341b2 |
| PPO | 52.3 | Not recorded | 693266f2d6b0dd33429341b2 |
| RFT | 49.8 | Not recorded | 693266f2d6b0dd33429341b2 |

### ReAct

Accuracy: 80.8. Result ID: 699988c85698665ec4fa3f0c.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6eb41e209131cea41992.md>)

[Implementation](<https://react-lm.github.io/>)

Recorded evaluation configuration: {"Paradigm":"Prompting","Base Model":"GPT5"}. This record does not assert independent reproduction.

### PPO

Accuracy: 73.4. Result ID: 699988c85698665ec4fa3f15.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63a01e209131ce9ffdaa.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"QwQ-32B"}. This record does not assert independent reproduction.

### NLAC

Accuracy: 72.7. Result ID: 699988c85698665ec4fa3f1a.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"QwQ-32B"}. This record does not assert independent reproduction.

### GRPO

Accuracy: 71.8. Result ID: 699988c85698665ec4fa3f16.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

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

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

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"QwQ-32B"}. This record does not assert independent reproduction.

### NLRL

Accuracy: 71.5. Result ID: 699988c85698665ec4fa3f17.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78d31e209131cea7a9ba.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"QwQ-32B"}. This record does not assert independent reproduction.

### NLQL

Accuracy: 71.2. Result ID: 699988c85698665ec4fa3f19.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"QwQ-32B","Notes":"ablation"}. This record does not assert independent reproduction.

### Self-Distillation

Accuracy: 70.1. Result ID: 699988c85698665ec4fa3f18.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"QwQ-32B","Notes":"ablation"}. This record does not assert independent reproduction.

### RFT

Accuracy: 69.4. Result ID: 699988c85698665ec4fa3f14.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"QwQ-32B"}. This record does not assert independent reproduction.

### Self-Distillation

Accuracy: 57.1. Result ID: 699988c85698665ec4fa3f11.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"Qwen2.5-7B-Instruct","Notes":"ablation"}. This record does not assert independent reproduction.

### NLQL

Accuracy: 56.4. Result ID: 699988c85698665ec4fa3f12.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"Qwen2.5-7B-Instruct","Notes":"ablation"}. This record does not assert independent reproduction.

### NLAC

Accuracy: 56.2. Result ID: 699988c85698665ec4fa3f13.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"Qwen2.5-7B-Instruct"}. This record does not assert independent reproduction.

### NLRL

Accuracy: 53.4. Result ID: 699988c85698665ec4fa3f10.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa78d31e209131cea7a9ba.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"Qwen2.5-7B-Instruct"}. This record does not assert independent reproduction.

### GRPO

Accuracy: 52.5. Result ID: 699988c85698665ec4fa3f0f.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

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

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

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"Qwen2.5-7B-Instruct"}. This record does not assert independent reproduction.

### PPO

Accuracy: 52.3. Result ID: 699988c85698665ec4fa3f0e.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa63a01e209131ce9ffdaa.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"Qwen2.5-7B-Instruct"}. This record does not assert independent reproduction.

### RFT

Accuracy: 49.8. Result ID: 699988c85698665ec4fa3f0d.

[Reporting paper](<https://www.sota2.com/research/paper/693266f2d6b0dd33429341b2.md>)

Recorded evaluation configuration: {"Paradigm":"Fine-tuning","Base Model":"Qwen2.5-7B-Instruct"}. This record does not assert independent reproduction.

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