Scientific Question Answering on GPQA (Pass@8)
79.29Pass@8MOTAB
Evaluation Results
| Method | Links | |
|---|---|---|
| MOTABStudent LLM=Qwen2.5-7B-Instruct, Teacher Model=Qwen3-32B2026.05 | 79.29 | |
| SKDStudent LLM=Qwen2.5-7B-Instruct, Teacher Model=Qwen3-32B, beta=1.02026.05 | 77.27 | |
| SKDStudent LLM=Qwen2.5-7B-Instruct, Teacher Model=Qwen3-32B, beta=0.82026.05 | 76.77 | |
| SKDStudent LLM=Qwen3-1.7B, Teacher Model=Qwen3-32B, beta=0.82026.05 | 73.74 | |
| MOTABStudent LLM=Qwen3-1.7B, Teacher Model=Qwen3-32B2026.05 | 73.74 | |
| MOTABStudent LLM=Llama3.2-3B-Instruct, Teacher Model=Qwen3-32B2026.05 | 73.74 | |
| SKDStudent LLM=Llama3.2-3B-Instruct, Teacher Model=Qwen3-32B, beta=1.02026.05 | 71.21 | |
| ImitKDStudent LLM=Qwen2.5-7B-Instruct, Teacher Model=Qwen3-32B2026.05 | 70.71 | |
| SKDStudent LLM=Qwen2.5-7B-Instruct, Teacher Model=Qwen3-32B, beta=0.62026.05 | 70.71 | |
| SFTStudent LLM=Qwen2.5-7B-Instruct, Teacher Model=Qwen3-32B2026.05 | 69.7 | |
| SFTStudent LLM=Qwen3-1.7B, Teacher Model=Qwen3-32B2026.05 | 68.18 | |
| SKDStudent LLM=Llama3.2-3B-Instruct, Teacher Model=Qwen3-32B, beta=0.82026.05 | 67.68 | |
| ImitKDStudent LLM=Llama3.2-3B-Instruct, Teacher Model=Qwen3-32B2026.05 | 66.67 | |
| SKDStudent LLM=Llama3.2-3B-Instruct, Teacher Model=Qwen3-32B, beta=0.62026.05 | 64.65 | |
| ImitKDStudent LLM=Qwen3-1.7B, Teacher Model=Qwen3-32B2026.05 | 63.13 | |
| SKDStudent LLM=Qwen3-1.7B, Teacher Model=Qwen3-32B, beta=1.02026.05 | 62.63 | |
| SKDStudent LLM=Qwen3-1.7B, Teacher Model=Qwen3-32B, beta=0.62026.05 | 61.62 | |
| RFTStudent LLM=Qwen3-1.7B, Teacher Model=Qwen3-32B2026.05 | 60.1 | |
| SFTStudent LLM=Llama3.2-3B-Instruct, Teacher Model=Qwen3-32B2026.05 | 58.59 | |
| RFTStudent LLM=Qwen2.5-7B-Instruct, Teacher Model=Qwen3-32B2026.05 | 56.06 | |
| RFTStudent LLM=Llama3.2-3B-Instruct, Teacher Model=Qwen3-32B2026.05 | 53.54 |