Mathematical Reasoning on AMC23 (Avg@16)
77.6Avg@16RTS
Evaluation results
| Method | Links | |
|---|---|---|
| RTSBackbone=Qwen3-4B (non-think)2026.01 | 77.6 | |
| Ours (Shorten Prefix)Backbone=Qwen3-4B (non-think)2026.01 | 74.5 | |
| SENTBase Model=DeepSeek-R1-Distill-Qwen-1.5B2025.12 | 70.16 | |
| GRPOBackbone=Qwen3-4B (non-think)2026.01 | 69.2 | |
| MixedBackbone=Qwen3-4B (non-think)2026.01 | 69.1 | |
| BaseBackbone=Qwen3-4B (non-think)2026.01 | 66.7 | |
| MixedBackbone=Qwen3-8B-Base2026.01 | 64.5 | |
| GRPOBackbone=Qwen3-8B-Base2026.01 | 63.4 | |
| Ours (Shorten Prefix)Backbone=Qwen3-8B-Base2026.01 | 63.3 | |
| RTSBackbone=Qwen3-8B-Base2026.01 | 62.8 | |
| Ours (Shorten Prefix)Backbone=Qwen3-4B-Base2026.01 | 60 | |
| GRPOBackbone=Qwen3-4B-Base2026.01 | 59.2 | |
| LENSbackbone=Qwen2.5-7B2026.01 | 56.41 | |
| RTSBackbone=Qwen3-4B-Base2026.01 | 54.1 | |
| DAPObackbone=Qwen2.5-7B2026.01 | 53.91 | |
| MixedBackbone=Qwen3-4B-Base2026.01 | 53.1 | |
| GRPOextendedbackbone=Qwen2.5-7B2026.01 | 51.41 | |
| GRPObackbone=Qwen2.5-7B2026.01 | 51.09 | |
| GRESObackbone=Qwen2.5-7B2026.01 | 48.28 | |
| OpenPangu-embedded-1B-v1.1Base Model=DeepSeek-R1-Distill-Qwen-1.5B2025.12 | 46.25 | |
| LENSbackbone=Qwen2.5-3B2026.01 | 38.91 | |
| GRESObackbone=Qwen2.5-3B2026.01 | 37.19 | |
| BaseBackbone=Qwen3-4B-Base2026.01 | 35.9 | |
| DAPObackbone=Qwen2.5-3B2026.01 | 34.06 | |
| GRPObackbone=Qwen2.5-3B2026.01 | 31.09 | |
| GRPOextendedbackbone=Qwen2.5-3B2026.01 | 27.81 | |
| BaseBackbone=Qwen3-8B-Base2026.01 | 26.4 | |
| Qwen2.5-7Bbackbone=Qwen2.5-7B2026.01 | 19.69 | |
| Qwen2.5-3Bbackbone=Qwen2.5-3B2026.01 | 12.97 |