LLM Evaluation on Shared (evaluation)
78Tie-aware AccuracyGrowLoop
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GrowLoopCategory=Ours2026.05 | 78 | 87 | 0.78 | |
| ICAICategory=Training-free rubric2026.05 | 58 | 85 | 0.75 | |
| OpenJudgeCategory=Training-free rubric2026.05 | 58 | 70 | 0.62 | |
| ICL (k=3)Category=No rubric2026.05 | 37 | 57 | 0.09 | |
| Arena-Hard PromptCategory=Manual rubric2026.05 | 27 | 43 | -0.17 | |
| Zero-shotCategory=No rubric2026.05 | 25 | 42 | -0.21 | |
| MT-Bench PromptCategory=Manual rubric2026.05 | 23 | 38 | -0.31 | |
| Skywork-Reward-V2Category=Reward Model2026.05 | 22 | 39 | -0.2 | |
| RM-R1Category=Reward Model2026.05 | 15 | 25 | -0.5 | |
| OpenRubric-JudgeCategory=Training-based rubric2026.05 | 14 | 24 | -0.49 |