## Evaluation Results

|                                 | AIME2024 | MATH500 | GPQA-Diamond | AIME2025 Part I |
|-|-|-|-|-|
| Bespoke-Stratos-32B            | 63.3     | 93.0    | 58.1         | -                |
| Sky-T1-32B                      | 43.3     | 82.4    | 56.8         | -                |
| DeepSeek-R1-Distill-Qwen-32B    | 66.7     | 89.8    | 61.1         | 53.3             |
| OpenThinker-32B                 | 66.0     | 90.6    | 61.6         | 53.3             |
| **Control-R-32B (Ours)**        | **70.0** | **93.2** | 61.1         | 55.0             |
| o1-preview                      | 40.0     | 81.4    | **75.2**     | **78.3**         |
| DeepSeek-R1                     | **79.8** | **97.3** | **71.5**     | 65.0             |

## Usage

### Prompt

```
query = '...'
control_fields = "\n<control> search_depth: 9; search_breadth: 9; error_detection: 9; error_correction: 9; strategy_switching: 9; correctness: 9; efficiency: 9; completeness: 9; coherence: 9; knowledge_accuracy: 9; clarity_of_steps: 9 <control/>"
query += control_fields + "\nPlease reason step by step, and put your final answer within \\boxed{}.",