Value Modeling on DAPO-Math-17k Qwen3-4B-Instruct-2507 policy (Held-out)
0.689Intra AUCV0
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| V0Protocol=Strict Generalization (Zero-Shot Transfer)2026.02 | 0.689 | 0.804 | 0.138 | |
| Vanilla Value ModelProtocol=Strict Generalization (Zero-Shot Transfer)2026.02 | 0.512 | 0.304 | 0.474 |