Reward Prediction on SIQA (out-of-domain)
76.89AccuracyMemReward
Evaluation Results
| Method | Links | |
|---|---|---|
| MemRewardBackbone=Qwen2.5-3B-Instruct2026.03 | 76.89 | |
| R1-OracleBackbone=Qwen2.5-3B-Instruct2026.03 | 76.89 | |
| R1-OracleBackbone=Qwen2.5-1.5B-Instruct2026.03 | 74.89 | |
| R1-pBackbone=Qwen2.5-3B-Instruct2026.03 | 74.67 | |
| MemRewardBackbone=Qwen2.5-1.5B-Instruct2026.03 | 74.44 | |
| R1-pBackbone=Qwen2.5-1.5B-Instruct2026.03 | 72.67 |