Reasoning on AIME 24
34.58Avg@32CraEG
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| CraEGTemperature=0.7, Top-p=0.95, Backbone=Qwen3-1.7B2026.01 | 34.58 | 56.16 | 44.76 | 12.52 | — | |
| Standard SamplingTemperature=0.7, Top-p=0.95, Backbone=Qwen3-1.7B2026.01 | 34.17 | 57.53 | 44.7 | 12.43 | — | |
| CraEGTemperature=1, Top-p=1, Backbone=Qwen3-1.7B2026.01 | 34.06 | 56.59 | 51.91 | 15.25 | — | |
| Standard SamplingTemperature=1, Top-p=1, Backbone=Qwen3-1.7B2026.01 | 33.75 | 55.07 | 51 | 15.17 | — | |
| MTModel Family=Qwen 2.5, Model Size=7B, Role=Teacher2026.04 | 28.13 | — | — | — | — | |
| MTModel Family=LLaMA 3, Model Size=8B, Role=Teacher2026.04 | 14.27 | — | — | — | — | |
| HPDModel Family=Qwen 2.5, Student Model=3B2026.04 | 13.75 | — | — | — | — | |
| SeqKDModel Family=Qwen 2.5, Student Model=3B2026.04 | 11.56 | — | — | — | — | |
| HPDModel Family=LLaMA 3, Student Model=3B2026.04 | 10.94 | — | — | — | — | |
| JSDModel Family=Qwen 2.5, Student Model=3B2026.04 | 10.31 | — | — | — | — | |
| SFTModel Family=Qwen 2.5, Student Model=3B2026.04 | 10.1 | — | — | — | — | |
| RKLDModel Family=Qwen 2.5, Student Model=3B2026.04 | 9.38 | — | — | — | — | |
| SeqKDModel Family=LLaMA 3, Student Model=3B2026.04 | 9.06 | — | — | — | — | |
| SFTModel Family=LLaMA 3, Student Model=3B2026.04 | 8.13 | — | — | — | — | |
| HPDModel Family=Qwen 2.5, Student Model=1.5B2026.04 | 7.71 | — | — | — | — | |
| RKLDModel Family=LLaMA 3, Student Model=3B2026.04 | 7.19 | — | — | — | — | |
| MSModel Family=LLaMA 3, Student Model=3B, Role=Student Baseline2026.04 | 7.08 | — | — | — | — | |
| MSModel Family=Qwen 2.5, Student Model=3B, Role=Student Baseline2026.04 | 6.67 | — | — | — | — | |
| JSDModel Family=LLaMA 3, Student Model=3B2026.04 | 6.67 | — | — | — | — | |
| JSDModel Family=Qwen 2.5, Student Model=1.5B2026.04 | 5.73 | — | — | — | — | |
| SeqKDModel Family=Qwen 2.5, Student Model=1.5B2026.04 | 5.31 | — | — | — | — | |
| RKLDModel Family=Qwen 2.5, Student Model=1.5B2026.04 | 5 | — | — | — | — | |
| SFTModel Family=Qwen 2.5, Student Model=1.5B2026.04 | 2.81 | — | — | — | — | |
| MSModel Family=Qwen 2.5, Student Model=1.5B, Role=Student Baseline2026.04 | 2.19 | — | — | — | — | |
| HPDModel Family=LLaMA 3, Student Model=1B2026.04 | 1.25 | — | — | — | — | |
| SFTModel Family=LLaMA 3, Student Model=1B2026.04 | 0.83 | — | — | — | — | |
| MSModel Family=LLaMA 3, Student Model=1B, Role=Student Baseline2026.04 | 0.73 | — | — | — | — | |
| SeqKDModel Family=LLaMA 3, Student Model=1B2026.04 | 0.42 | — | — | — | — | |
| RKLDModel Family=LLaMA 3, Student Model=1B2026.04 | 0.42 | — | — | — | — | |
| JSDModel Family=LLaMA 3, Student Model=1B2026.04 | 0.31 | — | — | — | — | |
| DeepSeek-LlamaBackbone=DeepSeek-Llama2025.05 | — | — | — | — | 73.3 | |
| DeepSeek-QwenBackbone=DeepSeek-Qwen2025.05 | — | — | — | — | 76.7 | |
| InfiniGenBackbone=DeepSeek-Llama2025.05 | — | — | — | — | 36.7 | |
| InfiniGenBackbone=DeepSeek-Qwen2025.05 | — | — | — | — | 60 | |
| PQCacheBackbone=DeepSeek-Llama2025.05 | — | — | — | — | 30 | |
| PQCacheBackbone=DeepSeek-Qwen2025.05 | — | — | — | — | 13.33 | |
| QuestBackbone=DeepSeek-Llama2025.05 | — | — | — | — | 46.7 | |
| QuestBackbone=DeepSeek-Qwen2025.05 | — | — | — | — | 46.7 | |
| RetroInferBackbone=DeepSeek-Llama2025.05 | — | — | — | — | 76.7 | |
| RetroInferBackbone=DeepSeek-Qwen2025.05 | — | — | — | — | 73.3 |