| Dataset Name | SOTA Method | Metric | Trend | ||
|---|---|---|---|---|---|
| RewardBench 2 | Scalar RM | Factuality88.2 | 21 | 1mo ago | |
| RewardBench | Spearman ρ0.542 | 20 | 2mo ago | ||
| RewardBench | SAVE | Accuracy93.9 | 12 | 1mo ago | |
| Reward Bench 2 (test) | Distribution-Calibrated Aggregation | RB2 Factuality MAE0.451 | 12 | 4mo ago | |
| internally annotated (test) | GDSO | PLCC0.762 | 11 | 1mo ago | |
| Alpaca-Eval | FiMi-RM | LC Win Rate72.59 | 8 | 1mo ago | |
| RewardBench (test) | Consistency | Kuiper1.65 | 8 | 4mo ago | |
| Text-Expert (test fraction 0.2) | Pair Accuracy80.6 | 7 | 29d ago | ||
| Text-Math fraction 0.2 (test) | Pair Accuracy73 | 7 | 29d ago | ||
| Text-Coding (test fraction 0.2) | Pair Accuracy80.9 | 7 | 29d ago | ||
| RewardBench | DPO | P-value0.001 | 7 | 2mo ago | |
| Meta-World Object OOD | FLORA | Process Alignment Correlation (ρ)0.81 | 5 | 2mo ago | |
| Meta-World Viewpoint OOD | FLORA | Process Alignment ρ0.88 | 5 | 2mo ago | |
| Meta-World Position OOD | FLORA | Process Alignment ρ0.85 | 5 | 2mo ago | |
| Meta-World (train) | FLORA | Procedural Alignment Correlation (ρ)0.97 | 5 | 2mo ago | |
| Arena-Hard RU | Qwen3-32B-RM | Best@8 Score92.69 | 5 | 4mo ago |
