Reward Modeling Accuracy on InfoBench
87.7AccuracyRubric-ARM-voting@5
Evaluation results
| Method | Links | |
|---|---|---|
| Rubric-ARM-voting@5voting=52026.02 | 87.7 | |
| RM-R1-32B (Qwen-2.5-Inst)Backbone=Qwen-2.5-Inst, Size=32B2026.02 | 86.1 | |
| Rubric-ARM2026.02 | 86.1 | |
| Gemini-2.5-Flash2026.02 | 85.6 | |
| RM-R1-14B (Qwen-2.5-Inst)Backbone=Qwen-2.5-Inst, Size=14B2026.02 | 85.5 | |
| RM-R1-32B (DeepSeek-Dist)Backbone=DeepSeek-Dist, Size=32B2026.02 | 85 | |
| RRM-32BSize=32B2026.02 | 84.4 | |
| Rubric-RM-voting@5voting=52026.02 | 83.8 | |
| RM-R1-14B (DeepSeek-Dist)Backbone=DeepSeek-Dist, Size=14B2026.02 | 82.4 | |
| API (Rubric+Judge)Rubric Model=GPT-4.1-Mini, Judge Model=Gemini-2.5-Flash Lite2026.02 | 82.2 | |
| Rubric-RM2026.02 | 80.8 | |
| Qwen-3-8B (Rubric+Judge)Backbone=Qwen-3-8B2026.02 | 74.6 | |
| API (direct Judge)Judge Model=Gemini-2.5-Flash Lite2026.02 | 72.9 | |
| RM-R1-7B (Qwen-2.5-Inst)Backbone=Qwen-2.5-Inst, Size=7B2026.02 | 71.3 | |
| RM-R1-7B (DeepSeek-Dist)Backbone=DeepSeek-Dist, Size=7B2026.02 | 70.3 | |
| RRM-7BSize=7B2026.02 | 68.2 | |
| JudgeLRM-7BSize=7B2026.02 | 62.7 |