Mathematical Reasoning on HMMT 2025 (test)
93.3AccuracyRSA
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| RSAModel 2=GPT-5 mini, Pairing Type=Heterogeneous2026.04 | 93.3 | 0.94 | 1 | |
| SQUEEZE EVOLVEModel 1=GPT-OSS-20B, Model 2=GPT-5 mini, p-value=30, Pairing Type=Heterogeneous2026.04 | 93.1 | 0.56 | 1.7 | |
| SQUEEZE EVOLVEModel 1=GPT-OSS-20B, Model 2=GPT-OSS-120B, p-value=10, Pairing Type=Homogeneous2026.04 | 92 | 0.25 | 1.6 | |
| RSAModel 2=GPT-OSS-120B, Pairing Type=Homogeneous2026.04 | 89.7 | 0.41 | 1 | |
| RA-RFTModel Scale=Qwen3-4B, Temperature=1.0, Maximum generation length=32,768, Sample count (k)=322026.06 | 47.3 | — | — | |
| GRPOModel Scale=Qwen3-4B, Temperature=1.0, Maximum generation length=32,768, Sample count (k)=322026.06 | 46.4 | — | — | |
| OPSDModel Scale=Qwen3-4B, Temperature=1.0, Maximum generation length=32,768, Sample count (k)=162026.06 | 45.2 | — | — | |
| Base (Instruct)Model Scale=Qwen3-4B, Temperature=1.0, Maximum generation length=32,768, Sample count (k)=322026.06 | 41.3 | — | — | |
| RA-RFTModel Scale=Qwen3-1.7B, Temperature=1.0, Maximum generation length=32,768, Sample count (k)=322026.06 | 28.2 | — | — | |
| GRPOModel Scale=Qwen3-1.7B, Temperature=1.0, Maximum generation length=32,768, Sample count (k)=322026.06 | 26.3 | — | — | |
| QuestAModel Scale=Qwen3-1.7B, Temperature=1.0, Maximum generation length=32,768, Sample count (k)=322026.06 | 26 | — | — | |
| OPSDModel Scale=Qwen3-1.7B, Temperature=1.0, Maximum generation length=32,768, Sample count (k)=162026.06 | 25 | — | — | |
| Base (Instruct)Model Scale=Qwen3-1.7B, Temperature=1.0, Maximum generation length=32,768, Sample count (k)=322026.06 | 23.4 | — | — |