Mathematical Reasoning on HMMT Feb 2024
19.6Average@16Ours (Shorten Prefix)
Evaluation Results
| Method | Links | |
|---|---|---|
| Ours (Shorten Prefix)Backbone=Qwen3-4B (non-think)2026.01 | 19.6 | |
| GRPOBackbone=Qwen3-4B (non-think)2026.01 | 16.7 | |
| MixedBackbone=Qwen3-4B (non-think)2026.01 | 13.7 | |
| RTSBackbone=Qwen3-4B (non-think)2026.01 | 13.7 | |
| Ours (Shorten Prefix)Backbone=Qwen3-4B-Base2026.01 | 11 | |
| BaseBackbone=Qwen3-4B (non-think)2026.01 | 10.6 | |
| RTSBackbone=Qwen3-8B-Base2026.01 | 9.5 | |
| GRPOBackbone=Qwen3-4B-Base2026.01 | 8.3 | |
| RTSBackbone=Qwen3-4B-Base2026.01 | 8 | |
| MixedBackbone=Qwen3-8B-Base2026.01 | 7.9 | |
| Ours (Shorten Prefix)Backbone=Qwen3-8B-Base2026.01 | 7.9 | |
| GRPOBackbone=Qwen3-8B-Base2026.01 | 6.4 | |
| MixedBackbone=Qwen3-4B-Base2026.01 | 6 | |
| BaseBackbone=Qwen3-4B-Base2026.01 | 2.3 | |
| BaseBackbone=Qwen3-8B-Base2026.01 | 1.7 |