Inference Latency (s) on AIME25
14.2Inference Latency (s)T2
Evaluation Results
| Method | Links | |
|---|---|---|
| T2Backbone=Llama3.1-8B2026.01 | 14.2 | |
| T2Backbone=Qwen2.5-7B2026.01 | 17.9 | |
| DSCBackbone=Qwen2.5-7B2026.01 | 25.5 | |
| Std-PTBackbone=Qwen2.5-7B2026.01 | 26.3 | |
| DSCBackbone=Llama3.1-8B2026.01 | 26.8 | |
| Std-PTBackbone=Llama3.1-8B2026.01 | 27.3 | |
| T2Backbone=Qwen3-4B2026.01 | 28.8 | |
| ESCBackbone=Qwen2.5-7B2026.01 | 38.1 | |
| DeepConfBackbone=Qwen2.5-7B2026.01 | 49.4 | |
| Std-PTBackbone=Qwen3-4B2026.01 | 51.4 | |
| ESCBackbone=Llama3.1-8B2026.01 | 66.6 | |
| ACBackbone=Qwen2.5-7B2026.01 | 74.4 | |
| ACBackbone=Llama3.1-8B2026.01 | 95.6 | |
| T2Backbone=Deepseek-R1-Distill-Qwen-7B2026.01 | 112.7 | |
| ESCBackbone=Qwen3-4B2026.01 | 115.6 | |
| ACBackbone=Qwen3-4B2026.01 | 123.9 | |
| DeepConfBackbone=Llama3.1-8B2026.01 | 142.5 | |
| DSCBackbone=Qwen3-4B2026.01 | 142.9 | |
| DeepConfBackbone=Qwen3-4B2026.01 | 193.3 | |
| Std-PTBackbone=Deepseek-R1-Distill-Qwen-7B2026.01 | 225.4 | |
| DSCBackbone=Deepseek-R1-Distill-Qwen-7B2026.01 | 450.8 | |
| ESCBackbone=Deepseek-R1-Distill-Qwen-7B2026.01 | 493.6 | |
| ACBackbone=Deepseek-R1-Distill-Qwen-7B2026.01 | 619.8 | |
| DeepConfBackbone=Deepseek-R1-Distill-Qwen-7B2026.01 | 804.7 |