Mathematical Reasoning on HMMT 2025 (Feb)
58Pass@1Pass@1
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Pass@1N=120, Backbone=GPT-OSS-20B2026.05 | 58 | — | — | — | — | — | — | |
| Pass@1N=24, Backbone=GPT-OSS-20B2026.05 | 57.9 | — | — | — | — | — | — | |
| Pass@1N=48, Backbone=GPT-OSS-20B2026.05 | 57.9 | — | — | — | — | — | — | |
| AOPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 38.33 | — | — | — | — | — | 52.42 | |
| OPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 34.17 | — | — | — | — | — | 50.86 | |
| ExOPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 34.17 | — | — | — | — | — | 50.22 | |
| GKDBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 34.17 | — | — | — | — | — | 49.67 | |
| OPDBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 32.92 | — | — | — | — | — | 47.65 | |
| ExOPDBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 32.08 | — | — | — | — | — | 47.3 | |
| AOPDBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 32.08 | — | — | — | — | — | 47.4 | |
| SFT CheckpointBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 31.04 | — | — | — | — | — | 44.19 | |
| AOPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 30.42 | — | — | — | — | — | 46.73 | |
| GKDBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 29.79 | — | — | — | — | — | 45.39 | |
| SFT CheckpointBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 27.08 | — | — | — | — | — | 38.97 | |
| GKDBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 25.83 | — | — | — | — | — | 40.21 | |
| SeqKDBackbone=Qwen3-8B-Base, SFT Warm-up steps=3K2026.05 | 25.42 | — | — | — | — | — | 43.8 | |
| SeqKDBackbone=Qwen3-4B-Base, SFT Warm-up steps=3K2026.05 | 24.58 | — | — | — | — | — | 38.01 | |
| SFT CheckpointBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 24.58 | — | — | — | — | — | 34.68 | |
| ExOPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 24.37 | — | — | — | — | — | 38.02 | |
| OPDBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 23.13 | — | — | — | — | — | 34.95 | |
| SeqKDBackbone=Qwen3-8B-Base, SFT Warm-up steps=1K2026.05 | 20.62 | — | — | — | — | — | 36.78 | |
| BoNN=24, Backbone=GPT-OSS-20B2026.05 | — | 91.1 | — | — | — | — | — | |
| BoNN=48, Backbone=GPT-OSS-20B2026.05 | — | 90.8 | — | — | — | — | — | |
| BoNN=120, Backbone=GPT-OSS-20B2026.05 | — | 90.6 | — | — | — | — | — | |
| JCN=24, Backbone=GPT-OSS-20B2026.05 | — | — | — | — | — | 96.5 | — | |
| JCN=48, Backbone=GPT-OSS-20B2026.05 | — | — | — | — | — | 98.2 | — | |
| JCN=120, Backbone=GPT-OSS-20B2026.05 | — | — | — | — | — | 99.8 | — | |
| KTN=24, Backbone=GPT-OSS-20B2026.05 | — | — | 21.3 | — | — | — | — | |
| KTN=48, Backbone=GPT-OSS-20B2026.05 | — | — | 27 | — | — | — | — | |
| KTN=120, Backbone=GPT-OSS-20B2026.05 | — | — | 38.5 | — | — | — | — | |
| SCN=24, Backbone=GPT-OSS-20B2026.05 | — | — | — | 92.2 | — | — | — | |
| SCN=48, Backbone=GPT-OSS-20B2026.05 | — | — | — | 93.9 | — | — | — | |
| SCN=120, Backbone=GPT-OSS-20B2026.05 | — | — | — | 94.3 | — | — | — | |
| WSCN=24, Backbone=GPT-OSS-20B2026.05 | — | — | — | — | 94.9 | — | — | |
| WSCN=48, Backbone=GPT-OSS-20B2026.05 | — | — | — | — | 97.6 | — | — | |
| WSCN=120, Backbone=GPT-OSS-20B2026.05 | — | — | — | — | 99.7 | — | — |