Math Reasoning on Math Reasoning Tasks Group 1 (Aggregate)
61.8MATH500 ScoreS1.1-3B + CGD
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| S1.1-3B + CGDBase Model=S1.1-3B, Fine-tuning Strategy=Critique-Guided Distillation (CGD), Teacher Model=LLaMA3.3-70B Instruct2025.05 | 61.8 | 27.9 | 82.5 | 23.1 | 35 | 46.1 | |
| S1.1-3B + Distilled SFTBase Model=S1.1-3B, Fine-tuning Strategy=Distilled SFT2025.05 | 60.6 | 22.1 | 83.1 | 20.4 | 22.5 | 41.7 | |
| S1.1-3B + SFTBase Model=S1.1-3B, Fine-tuning Strategy=SFT2025.05 | 55.4 | 18.8 | 76.8 | 19.6 | 30 | 40.1 | |
| LLaMA3.1-8B Instruct + CFT*Base Model=LLaMA3.1-8B Instruct, Fine-tuning Strategy=CFT*, Teacher Model=GPT-4o2025.05 | 54.8 | 33.1 | 86.2 | 18.2 | 25 | 43.5 | |
| LLaMA3.1-8B Instruct + CGDBase Model=LLaMA3.1-8B Instruct, Fine-tuning Strategy=Critique-Guided Distillation (CGD), Teacher Model=LLaMA3.3-70B Instruct2025.05 | 54.2 | 33.6 | 85.7 | 23.7 | 37.5 | 46.9 | |
| S1.1-3BBase Model=S1.1-3B2025.05 | 54 | 16.9 | 76.8 | 20.6 | 30 | 35.4 | |
| LLaMA3.1-8B Instruct + Distilled SFTBase Model=LLaMA3.1-8B Instruct, Fine-tuning Strategy=Distilled SFT2025.05 | 53.4 | 32.7 | 85.3 | 19.6 | 27.5 | 43.7 | |
| LLaMA3.1-8B Instruct + CFTBase Model=LLaMA3.1-8B Instruct, Fine-tuning Strategy=CFT, Teacher Model=LLaMA3.3-70B Instruct2025.05 | 51.8 | 32.7 | 84.8 | 15.7 | 22.5 | 41.5 | |
| LLaMA3.1-8B InstructBase Model=LLaMA3.1-8B Instruct2025.05 | 50.6 | 33.5 | 85.3 | 14.5 | 22.5 | 41.3 | |
| S1.1-3B + CFTBase Model=S1.1-3B, Fine-tuning Strategy=CFT, Teacher Model=LLaMA3.3-70B Instruct2025.05 | 49.6 | 21 | 77.3 | 19.3 | 27.5 | 38.9 | |
| LLaMA3.1-8B Instruct + SFTBase Model=LLaMA3.1-8B Instruct, Fine-tuning Strategy=SFT2025.05 | 41.2 | 24.6 | 80.7 | 10.8 | 20 | 35.5 |