Mathematical Reasoning on OlympiadBench (first 40 problems)
10Time (minutes)DiffAdapt
Evaluation results
| Method | Links | |
|---|---|---|
| DiffAdaptModel=Qwen3-4B, Backend=vLLM, Batch size=10, Hardware=single A800 GPU, Max token limit=32K, Temperature=0.62025.10 | 10 | |
| DEERModel=Qwen3-4B, Backend=vLLM, Batch size=10, Hardware=single A800 GPU, Max token limit=32K, Temperature=0.6, DEER think threshold=0.92025.10 | 57 | |
| vLLM BaselineModel=Qwen3-4B, Backend=vLLM, Batch size=10, Hardware=single A800 GPU, Max token limit=32K, Temperature=0.62025.10 | 64 |