Mathematical Reasoning on AIME24 (Accuracy, C_mem, C_time)
22C_mem RatioT2
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| T2Backbone model=Llama3.1-8B2026.01 | 22 | 43 | 9.34 | |
| DSCBackbone model=Qwen2.5-7B2026.01 | 27 | 183 | 15.52 | |
| DSCBackbone model=Llama3.1-8B2026.01 | 27 | 163 | 7.78 | |
| ESCBackbone model=Qwen2.5-7B2026.01 | 32 | 96 | 16.72 | |
| ESCBackbone model=Llama3.1-8B2026.01 | 33 | 164 | 9.28 | |
| ACBackbone model=Llama3.1-8B2026.01 | 34 | 356 | 8.85 | |
| T2Backbone model=Deepseek-R1-Distill-Qwen-7B2026.01 | 34 | 71 | 79.3 | |
| ACBackbone model=Deepseek-R1-Distill-Qwen-7B2026.01 | 44 | 250 | 78.64 | |
| T2Backbone model=Qwen2.5-7B2026.01 | 49 | 82 | 17.04 | |
| ESCBackbone model=Deepseek-R1-Distill-Qwen-7B2026.01 | 53 | 219 | 78.39 | |
| ESCBackbone model=Qwen3-4B2026.01 | 55 | 193 | 56.34 | |
| DSCBackbone model=Deepseek-R1-Distill-Qwen-7B2026.01 | 56 | 358 | 73.75 | |
| T2Backbone model=Qwen3-4B2026.01 | 57 | 79 | 57.01 | |
| DeepConfBackbone model=Llama3.1-8B2026.01 | 67 | 402 | 7.32 | |
| DSCBackbone model=Qwen3-4B2026.01 | 72 | 221 | 56.18 | |
| ACBackbone model=Qwen2.5-7B2026.01 | 74 | 346 | 16.65 | |
| DeepConfBackbone model=Deepseek-R1-Distill-Qwen-7B2026.01 | 83 | 498 | 79.58 | |
| ACBackbone model=Qwen3-4B2026.01 | 84 | 312 | 56.81 | |
| Std-PT.Backbone model=Qwen2.5-7B2026.01 | 100 | 100 | 16.72 | |
| Std-PT.Backbone model=Llama3.1-8B2026.01 | 100 | 100 | 9.15 | |
| Std-PT.Backbone model=Deepseek-R1-Distill-Qwen-7B2026.01 | 100 | 100 | 79.11 | |
| Std-PT.Backbone model=Qwen3-4B2026.01 | 100 | 100 | 57.21 | |
| DeepConfBackbone model=Qwen2.5-7B2026.01 | 113 | 316 | 15.08 | |
| DeepConfBackbone model=Qwen3-4B2026.01 | 119 | 463 | 57.3 |