Reward Model Evaluation on Text-Math fraction 0.2 (test)
73Pair AccuracyArena BT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Arena BTReward Model=proprietary RM2026.06 | 73 | 0.571 | 1 | 0.86 | |
| Arena Avg RewardReward Model=proprietary RM2026.06 | 73 | 0.571 | 1 | 0.86 | |
| DUALEVALReward Model=proprietary RM2026.06 | 68.2 | 0.99 | 0.734 | 0.703 | |
| Arena BTReward Model=Skywork-Reward-V2-Qwen3-8B2026.06 | 66.7 | 0.457 | 0.624 | 0.536 | |
| Arena Avg RewardReward Model=Skywork-Reward-V2-Qwen3-8B2026.06 | 66.1 | 0.445 | 0.639 | 0.503 | |
| Static 2PL2026.06 | 64.1 | 1 | 0.571 | 0.682 | |
| DUALEVALReward Model=Skywork-Reward-V2-Qwen3-8B2026.06 | 60 | 0.984 | 0.625 | 0.692 |