Reasoning on GPQA (avg.@8)
72Avg@8DeepSeek-Qwen
Evaluation Results
| Method | Links | |
|---|---|---|
| DeepSeek-QwenBackbone=DeepSeek-Qwen2025.05 | 72 | |
| DeepSeek-LlamaBackbone=DeepSeek-Llama2025.05 | 68 | |
| RetroInferBackbone=DeepSeek-Llama2025.05 | 68 | |
| RetroInferBackbone=DeepSeek-Qwen2025.05 | 68 | |
| InfiniGenBackbone=DeepSeek-Qwen2025.05 | 66 | |
| InfiniGenBackbone=DeepSeek-Llama2025.05 | 62 | |
| QuestBackbone=DeepSeek-Qwen2025.05 | 62 | |
| PQCacheBackbone=DeepSeek-Llama2025.05 | 58 | |
| QuestBackbone=DeepSeek-Llama2025.05 | 50 | |
| PQCacheBackbone=DeepSeek-Qwen2025.05 | 46 | |
| MTModel Family=Qwen 2.5, Model Size=7B2026.04 | 43.43 | |
| HPDModel Family=Qwen 2.5, Student Model=3B2026.04 | 31.31 | |
| HPDModel Family=LLaMA 3, Student Model=3B2026.04 | 27.15 |