# Reward Modeling on RewardBench English

Source: [Reward Modeling on RewardBench English](<https://www.sota2.com/research/sota/reward-modeling-on-rewardbench-english>)

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

Data updated: 2026-03-12T07:53:43.930Z

Evaluation of model performance on the English subset of the RewardBench dataset specifically for the Reward Modeling task, measured using 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 | Reward Modeling |
| Dataset (including any stated split) | RewardBench English |
| Metric | Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 14 of 14 |
| 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/reward-modeling-on-rewardbench-english.md?metric=Accuracy>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| DIBJudge-Qwen3-8B | 91.01 | Not recorded | 69b24878d030084c22e2f85c |
| DIBJudge-Qwen3-4B | 90.32 | Not recorded | 69b24878d030084c22e2f85c |
| mR3-Qwen3-8B | 90.1 | Not recorded | 69b24878d030084c22e2f85c |
| mR3-Qwen3-4B | 89.75 | Not recorded | 69b24878d030084c22e2f85c |
| Nemotron-Multi-49B | 89.71 | Not recorded | 69b24878d030084c22e2f85c |
| Gemini-2.5-Flash | 88.83 | Not recorded | 69b24878d030084c22e2f85c |
| Qwen3-8B | 88.81 | Not recorded | 69b24878d030084c22e2f85c |
| Think-as-Locals 7B | 88.79 | Not recorded | 69b24878d030084c22e2f85c |
| Qwen3-4B | 87.54 | Not recorded | 69b24878d030084c22e2f85c |
| GPT-4o | 85.96 | Not recorded | 69b24878d030084c22e2f85c |
| Qwen2.5-7B-Instruct | 78.59 | Not recorded | 69b24878d030084c22e2f85c |
| M-PROMETHEUS 7B | 76.69 | Not recorded | 69b24878d030084c22e2f85c |
| M-PROMETHEUS 3B | 69.79 | Not recorded | 69b24878d030084c22e2f85c |
| Qwen2.5-3B-Instruct | 68.99 | Not recorded | 69b24878d030084c22e2f85c |

### DIBJudge-Qwen3-8B

Accuracy: 91.01. Result ID: 69b270606dbb7075b95ccaa4.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

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

### DIBJudge-Qwen3-4B

Accuracy: 90.32. Result ID: 69b270606dbb7075b95ccaa3.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-4B"}. This record does not assert independent reproduction.

### mR3-Qwen3-8B

Accuracy: 90.1. Result ID: 69b270606dbb7075b95ccaa1.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bc51e209131cea8a58d.md>)

Recorded evaluation configuration: {"Model Category":"Multilingual Open Reward Models"}. This record does not assert independent reproduction.

### mR3-Qwen3-4B

Accuracy: 89.75. Result ID: 69b270606dbb7075b95ccaa0.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7bc51e209131cea8a58d.md>)

Recorded evaluation configuration: {"Model Category":"Multilingual Open Reward Models"}. This record does not assert independent reproduction.

### Nemotron-Multi-49B

Accuracy: 89.71. Result ID: 69b270606dbb7075b95cca9d.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa706f1e209131cea4af71.md>)

[Implementation](<https://github.com/i-Eval/FairEval>)

Recorded evaluation configuration: {"Model Category":"Multilingual Open Reward Models"}. This record does not assert independent reproduction.

### Gemini-2.5-Flash

Accuracy: 88.83. Result ID: 69b270606dbb7075b95cca98.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7c311e209131cea8cef5.md>)

Recorded evaluation configuration: {"Model Category":"Proprietary Models"}. This record does not assert independent reproduction.

### Qwen3-8B

Accuracy: 88.81. Result ID: 69b270606dbb7075b95cca9c.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

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

Recorded evaluation configuration: {"Model Category":"General Open Models"}. This record does not assert independent reproduction.

### Think-as-Locals 7B

Accuracy: 88.79. Result ID: 69b270606dbb7075b95ccaa2.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa7ba81e209131cea89b47.md>)

Recorded evaluation configuration: {"Model Category":"Multilingual Open Reward Models"}. This record does not assert independent reproduction.

### Qwen3-4B

Accuracy: 87.54. Result ID: 69b270606dbb7075b95cca9b.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

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

Recorded evaluation configuration: {"Model Category":"General Open Models"}. This record does not assert independent reproduction.

### GPT-4o

Accuracy: 85.96. Result ID: 69b270606dbb7075b95cca97.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75681e209131cea67a98.md>)

Recorded evaluation configuration: {"Model Category":"Proprietary Models"}. This record does not assert independent reproduction.

### Qwen2.5-7B-Instruct

Accuracy: 78.59. Result ID: 69b270606dbb7075b95cca9a.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa767e1e209131cea6d6a7.md>)

Recorded evaluation configuration: {"Model Category":"General Open Models"}. This record does not assert independent reproduction.

### M-PROMETHEUS 7B

Accuracy: 76.69. Result ID: 69b270606dbb7075b95cca9f.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77e71e209131cea75e93.md>)

Recorded evaluation configuration: {"Model Category":"Multilingual Open Reward Models"}. This record does not assert independent reproduction.

### M-PROMETHEUS 3B

Accuracy: 69.79. Result ID: 69b270606dbb7075b95cca9e.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa77e71e209131cea75e93.md>)

Recorded evaluation configuration: {"Model Category":"Multilingual Open Reward Models"}. This record does not assert independent reproduction.

### Qwen2.5-3B-Instruct

Accuracy: 68.99. Result ID: 69b270606dbb7075b95cca99.

[Reporting paper](<https://www.sota2.com/research/paper/69b24878d030084c22e2f85c.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa767e1e209131cea6d6a7.md>)

Recorded evaluation configuration: {"Model Category":"General Open Models"}. This record does not assert independent reproduction.

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