# Instruction on IFEval (avg@32 accuracy)

Source: [Instruction on IFEval (avg@32 accuracy)](<https://www.sota2.com/research/sota/instruction-on-ifeval-avg-at-32-accuracy>)

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

Data updated: 2026-05-28T08:54:10.833Z

Evaluation of model performance on the Instruction task using the IFEval dataset, measured specifically by avg@32 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 | Instruction |
| Dataset (including any stated split) | IFEval |
| Metric | Avg@32 Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 16 of 16 |
| 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@32 Accuracy](<https://www.sota2.com/research/sota/instruction-on-ifeval-avg-at-32-accuracy.md?metric=Avg%4032%20Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Avg@32 Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| IBTPO | 46.2 | Not recorded | 6a17cc00dbd7150506f81233 |
| GRPO w/ Clip-higher | 45.4 | Not recorded | 6a17cc00dbd7150506f81233 |
| IBRO | 44.7 | Not recorded | 6a17cc00dbd7150506f81233 |
| GRPO w/ Entropy Reg† | 44.2 | Not recorded | 6a17cc00dbd7150506f81233 |
| Initial Model | 43.8 | Not recorded | 6a17cc00dbd7150506f81233 |
| TreeRL | 42.5 | Not recorded | 6a17cc00dbd7150506f81233 |
| TreePO | 42.2 | Not recorded | 6a17cc00dbd7150506f81233 |
| Vanilla GRPO | 42 | Not recorded | 6a17cc00dbd7150506f81233 |
| IBTPO | 26.9 | Not recorded | 6a17cc00dbd7150506f81233 |
| IBRO | 25.4 | Not recorded | 6a17cc00dbd7150506f81233 |
| Vanilla GRPO | 24 | Not recorded | 6a17cc00dbd7150506f81233 |
| GRPO w/ Clip-higher | 23.7 | Not recorded | 6a17cc00dbd7150506f81233 |
| GRPO w/ Entropy Reg† | 23.7 | Not recorded | 6a17cc00dbd7150506f81233 |
| TreeRL | 23.5 | Not recorded | 6a17cc00dbd7150506f81233 |
| TreePO | 23.2 | Not recorded | 6a17cc00dbd7150506f81233 |
| Initial Model | 21.9 | Not recorded | 6a17cc00dbd7150506f81233 |

### IBTPO

Avg@32 Accuracy: 46.2. Result ID: 6a180298fbc373592c11dc41.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-8B-Base"}. This record does not assert independent reproduction.

### GRPO w/ Clip-higher

Avg@32 Accuracy: 45.4. Result ID: 6a180298fbc373592c11dc3c.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-8B-Base"}. This record does not assert independent reproduction.

### IBRO

Avg@32 Accuracy: 44.7. Result ID: 6a180298fbc373592c11dc3e.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a351e209131cea81c6d.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-8B-Base"}. This record does not assert independent reproduction.

### GRPO w/ Entropy Reg†

Avg@32 Accuracy: 44.2. Result ID: 6a180298fbc373592c11dc3d.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-8B-Base"}. This record does not assert independent reproduction.

### Initial Model

Avg@32 Accuracy: 43.8. Result ID: 6a180298fbc373592c11dc3a.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

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

Recorded evaluation configuration: {"Base Model":"Qwen3-8B-Base"}. This record does not assert independent reproduction.

### TreeRL

Avg@32 Accuracy: 42.5. Result ID: 6a180298fbc373592c11dc3f.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa796f1e209131cea7d97e.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-8B-Base"}. This record does not assert independent reproduction.

### TreePO

Avg@32 Accuracy: 42.2. Result ID: 6a180298fbc373592c11dc40.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7ae51e209131cea85c12.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-8B-Base"}. This record does not assert independent reproduction.

### Vanilla GRPO

Avg@32 Accuracy: 42. Result ID: 6a180298fbc373592c11dc3b.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

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

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

Recorded evaluation configuration: {"Base Model":"Qwen3-8B-Base"}. This record does not assert independent reproduction.

### IBTPO

Avg@32 Accuracy: 26.9. Result ID: 6a180298fbc373592c11dc39.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-1.7B-Base"}. This record does not assert independent reproduction.

### IBRO

Avg@32 Accuracy: 25.4. Result ID: 6a180298fbc373592c11dc36.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7a351e209131cea81c6d.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-1.7B-Base"}. This record does not assert independent reproduction.

### Vanilla GRPO

Avg@32 Accuracy: 24. Result ID: 6a180298fbc373592c11dc33.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

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

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

Recorded evaluation configuration: {"Base Model":"Qwen3-1.7B-Base"}. This record does not assert independent reproduction.

### GRPO w/ Clip-higher

Avg@32 Accuracy: 23.7. Result ID: 6a180298fbc373592c11dc34.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-1.7B-Base"}. This record does not assert independent reproduction.

### GRPO w/ Entropy Reg†

Avg@32 Accuracy: 23.7. Result ID: 6a180298fbc373592c11dc35.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-1.7B-Base"}. This record does not assert independent reproduction.

### TreeRL

Avg@32 Accuracy: 23.5. Result ID: 6a180298fbc373592c11dc37.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa796f1e209131cea7d97e.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-1.7B-Base"}. This record does not assert independent reproduction.

### TreePO

Avg@32 Accuracy: 23.2. Result ID: 6a180298fbc373592c11dc38.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7ae51e209131cea85c12.md>)

Recorded evaluation configuration: {"Base Model":"Qwen3-1.7B-Base"}. This record does not assert independent reproduction.

### Initial Model

Avg@32 Accuracy: 21.9. Result ID: 6a180298fbc373592c11dc32.

[Reporting paper](<https://www.sota2.com/research/paper/6a17cc00dbd7150506f81233.md>)

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

Recorded evaluation configuration: {"Base Model":"Qwen3-1.7B-Base"}. This record does not assert independent reproduction.

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