Mathematical Reasoning on AIME 2025 (VAcc, RAcc, ΔAcc, TFR)
87Verification AccuracyQwen3-235B-Thinking
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Qwen3-235B-ThinkingRewriter=GPT-5, Verifier=GPT-52025.06 | 87 | 67 | 20 | 23 | |
| DeepSeek-R1-Qwen3-8BRewriter=GPT-5, Verifier=GPT-52025.06 | 80 | 40 | 40 | 50 | |
| Qwen3-32BRewriter=GPT-5, Verifier=GPT-52025.06 | 50 | 20 | 30 | 60 |