Reward Modeling Evaluation on Reward Bench Ties 2
91.8Pairwise AccuracyDistribution-Calibrated Aggregation
Evaluation Results
| Method | Links | |
|---|---|---|
| Distribution-Calibrated Aggregationn=12, Judge LLM=gemini-2.5-flash2025.12 | 91.8 | |
| Distribution-Calibrated Aggregationn=4, Judge LLM=gemini-2.5-flash2025.12 | 90.5 | |
| SCn=4, Judge LLM=gemini-2.5-flash2025.12 | 84.4 | |
| SCn=12, Judge LLM=gemini-2.5-flash2025.12 | 84.2 | |
| CI-SCn=12, Judge LLM=gemini-2.5-flash2025.12 | 83.4 | |
| Soft-SCn=4, Judge LLM=gemini-2.5-flash2025.12 | 82.3 | |
| Soft-SCn=12, Judge LLM=gemini-2.5-flash2025.12 | 82.2 | |
| CI-SCn=4, Judge LLM=gemini-2.5-flash2025.12 | 82.2 | |
| GSCn=12, Judge LLM=gemini-2.5-flash2025.12 | 80.4 | |
| GSCn=4, Judge LLM=gemini-2.5-flash2025.12 | 79.2 | |
| USCn=12, Judge LLM=gemini-2.5-flash2025.12 | 77.9 | |
| USCn=4, Judge LLM=gemini-2.5-flash2025.12 | 77.3 |