Value Modeling on DAPO-Math-17k Qwen2.5-7B-Instruct policy (Held-out)
0.693Intra AUCV0
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| V0Protocol=Strict Generalization (Zero-Shot Transfer)2026.02 | 0.693 | 0.84 | 0.165 | |
| Vanilla Value ModelProtocol=Strict Generalization (Zero-Shot Transfer)2026.02 | 0.527 | 0.507 | 0.583 |