Listwise Judging on RewardBench listwise 2
58.1IF ScoreTIR-Judge-Distill
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| TIR-Judge-DistillParameters=8B2025.10 | 58.1 | 72.7 | 63.8 | 81.4 | 82 | 71.6 | |
| Gemini-2.5-flash-Preview2025.10 | 55.3 | 81.1 | 65.7 | 86.7 | 90.9 | 75.9 | |
| TIR-Judge-DistillParameters=4B2025.10 | 55 | 78.1 | 55.8 | 75 | 73.1 | 67.3 | |
| TIR-Judge-ZeroParameters=4B2025.10 | 47.5 | 86.4 | 59.3 | 85.2 | 62.9 | 68.3 | |
| TIR-Judge-ZeroParameters=8B2025.10 | 45.6 | 84.1 | 64.8 | 89.5 | 82.7 | 73.4 | |
| Claude-Opusversion=42025.10 | 41.9 | 74.9 | 82.7 | 86.2 | 89.5 | 76.5 | |
| GPT-4.1-mini2025.10 | 41.2 | 72.1 | 60.8 | 73.5 | 72.6 | 65.7 | |
| GPT-4.12025.10 | 39.7 | 65.2 | 82.9 | 73.4 | 87.3 | 69.7 | |
| Claude-Sonnetversion=42025.10 | 35.9 | 70.5 | 76.1 | 76 | 89.1 | 69.5 | |
| GPT-4o2025.10 | 33.1 | 62.3 | 56.8 | 72.9 | 86.2 | 64.9 |