Mathematical Reasoning on OlympiadBench (Accuracy, Average Length)
59.85AccuracyGraph-Based Chain-of-Thought Pruning
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Graph-Based Chain-of-Thought PruningModel Scale=7B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 59.85 | 3,786 | |
| AdaptThinkModel Scale=7B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 58.76 | 5,011 | |
| BaseModel Scale=7B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 56.77 | 5,252 | |
| EfficientReasoningModel Scale=7B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 56.52 | 3,920 | |
| OREAL-7BModel Scale=7B, Method Variant=Open-Source R1-Style2026.04 | 56.44 | 5,164 | |
| Light-R1-DS-7BModel Scale=7B, Method Variant=Open-Source R1-Style2026.04 | 54.74 | 5,306 | |
| O1-PrunerModel Scale=7B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 54.15 | 5,695 | |
| TokenSkipModel Scale=7B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 52.94 | 5,795 | |
| AReaL-boba-RL-7BModel Scale=7B, Method Variant=Open-Source R1-Style2026.04 | 52.91 | 6,037 | |
| Skywork-OR1-7BModel Scale=7B, Method Variant=Open-Source R1-Style2026.04 | 51.41 | 6,075 | |
| Graph-Based Chain-of-Thought PruningModel Scale=1.5B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 49.78 | 3,221 | |
| EfficientReasoningModel Scale=1.5B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 49.48 | 4,032 | |
| AdaptThinkModel Scale=1.5B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 46.67 | 2,019 | |
| BaseModel Scale=1.5B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 43.89 | 5,632 | |
| TokenSkipModel Scale=1.5B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 38.63 | 6,692 | |
| O1-PrunerModel Scale=1.5B, Backbone=DeepSeek-R1-Distill-Qwen2026.04 | 38.56 | 6,645 |