Pairwise Accuracy on RubricBench 1.0 (test)
81.8Pairwise Accuracy (Chat)SVR
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| SVRModel=OSS2026.06 | 81.8 | 85.5 | 82.4 | 83.4 | 83.8 | 82.8 | |
| Oracle RubricsModel=OSS2026.06 | 80.6 | 85.5 | 83.6 | 86.4 | 92.5 | 83.1 | |
| RRDModel=OSS2026.06 | 65.3 | 64 | 56.8 | 53.8 | 52.8 | 57.6 | |
| Scalar RMModel=Skywork2026.06 | 63 | 69.4 | 58 | 52 | 77.5 | 61 | |
| Self-Generated RubricsModel=GPT-5.42026.06 | 55.9 | 71 | 65.6 | 60.1 | 51.2 | 60.3 | |
| Self-Generated RubricsModel=Gemini-3-Flash2026.06 | 55.2 | 70.2 | 65.6 | 60.9 | 50 | 60.1 | |
| Self-Generated RubricsModel=OSS2026.06 | 54.7 | 69.4 | 64.4 | 59 | 48.8 | 59 | |
| Direct pairwise judgeModel=GPT-5.42026.06 | 53.1 | 66.1 | 69.6 | 64.9 | 56.3 | 61.1 | |
| Rubric-ARMModel=RubricARM-8B2026.06 | 51 | 43.6 | 51.6 | 50.2 | 28.8 | 48.6 | |
| Direct pairwise judgeModel=Gemini-3-Flash2026.06 | 50 | 69.4 | 66.8 | 63.1 | 36.3 | 57.9 | |
| Rubric-ARMModel=RubricARM-8B / OSS2026.06 | 48.3 | 58.1 | 61.2 | 54.2 | 25 | 52 | |
| Direct pairwise judgeModel=OSS2026.06 | 44.6 | 69.4 | 60.8 | 58.7 | 60 | 55.2 | |
| Generative RMModel=RM-R12026.06 | 43.1 | 39.5 | 50.4 | 47.2 | 53.8 | 46 |