Reward Model Evaluation on Text-Expert (test fraction 0.2)
80.6Pair AccuracyArena BT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Arena BTReward Model=proprietary RM2026.06 | 80.6 | 0.695 | 1 | 0.874 | |
| Arena Avg RewardReward Model=proprietary RM2026.06 | 80.6 | 0.695 | 1 | 0.874 | |
| DUALEVALReward Model=proprietary RM2026.06 | 78 | 0.893 | 0.94 | 0.833 | |
| Arena BTReward Model=Skywork-Reward-V2-Qwen3-8B2026.06 | 77.9 | 0.22 | 0.35 | 0.247 | |
| Arena Avg RewardReward Model=Skywork-Reward-V2-Qwen3-8B2026.06 | 77.7 | 0.228 | 0.354 | 0.222 | |
| Static 2PL2026.06 | 70.1 | 1 | 0.695 | 0.701 | |
| DUALEVALReward Model=Skywork-Reward-V2-Qwen3-8B2026.06 | 61.2 | 0.963 | 0.724 | 0.657 |