Question Answering on MedXpertQA standard (test)
41.7AccuracyGPT-4.1
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-4.1Approach=Verbalized Conf.2025.11 | 41.7 | 0.551 | 0.559 | 0.564 | |
| GPT-4.1Approach=Verbalized Top-k2025.11 | 40.4 | 0.583 | 0.399 | 0.402 | |
| GPT-4.1Approach=Verbalized Distrib.2025.11 | 40 | 0.656 | 0.309 | 0.29 | |
| GPT-4.1Approach=Standard2025.11 | 39.7 | — | — | — | |
| DeepSeek-V3Approach=Verbalized Conf.2025.11 | 31.3 | 0.548 | 0.542 | 0.573 | |
| DeepSeek-V3Approach=Verbalized Top-k2025.11 | 30.9 | 0.55 | 0.434 | 0.467 | |
| DeepSeek-V3Approach=Standard2025.11 | 30.7 | — | — | — | |
| DeepSeek-V3Approach=Verbalized Distrib.2025.11 | 29.7 | 0.574 | 0.302 | 0.275 | |
| Qwen3-30B-A3B-InstructApproach=Verbalized Conf.2025.11 | 25.8 | 0.565 | 0.672 | 0.689 | |
| Qwen3-30B-A3B-InstructApproach=Verbalized Top-k2025.11 | 25.2 | 0.525 | 0.658 | 0.683 | |
| Qwen3-30B-A3B-InstructApproach=Verbalized Distrib.2025.11 | 25.1 | 0.582 | 0.339 | 0.359 | |
| Qwen3-30B-A3B-InstructApproach=Standard2025.11 | 24.7 | — | — | — | |
| Qwen3-30B-A3B-InstructApproach=Logit2025.11 | 24.7 | 0.508 | 0.734 | 0.732 | |
| Qwen3-30B-A3B-InstructApproach=p(True)2025.11 | 24.7 | 0.51 | 0.734 | 0.736 | |
| Baichuan-M2-32B-GPTQ-INT4Approach=Verbalized Conf.2025.11 | 21.1 | 0.516 | 0.608 | 0.66 | |
| Baichuan-M2-32B-GPTQ-INT4Approach=Standard2025.11 | 20.7 | — | — | — | |
| Baichuan-M2-32B-GPTQ-INT4Approach=Logit2025.11 | 20.7 | 0.481 | 0.788 | 0.789 | |
| Baichuan-M2-32B-GPTQ-INT4Approach=p(True)2025.11 | 20.7 | 0.459 | 0.57 | 0.617 | |
| Baichuan-M2-32B-GPTQ-INT4Approach=Verbalized Distrib.2025.11 | 20.7 | 0.539 | 0.303 | 0.345 | |
| Mistral-3.2-24B-InstructApproach=Standard2025.11 | 20 | — | — | — | |
| Mistral-3.2-24B-InstructApproach=Logit2025.11 | 20 | 0.526 | 0.753 | 0.763 | |
| Mistral-3.2-24B-InstructApproach=p(True)2025.11 | 20 | 0.539 | 0.52 | 0.576 | |
| Mistral-3.2-24B-InstructApproach=Verbalized Distrib.2025.11 | 20 | 0.549 | 0.39 | 0.444 | |
| Baichuan-M2-32B-GPTQ-INT4Approach=Verbalized Top-k2025.11 | 18.7 | 0.509 | 0.526 | 0.607 | |
| Mistral-3.2-24B-InstructApproach=Verbalized Top-k2025.11 | 18.6 | 0.556 | 0.514 | 0.6 | |
| Mistral-3.2-24B-InstructApproach=Verbalized Conf.2025.11 | 18.1 | 0.536 | 0.64 | 0.695 | |
| Qwen3-4B-InstructApproach=Standard2025.11 | 17.6 | — | — | — | |
| Qwen3-4B-InstructApproach=Logit2025.11 | 17.6 | 0.505 | 0.81 | 0.808 | |
| Qwen3-4B-InstructApproach=p(True)2025.11 | 17.6 | 0.497 | 0.534 | 0.572 | |
| Qwen3-4B-InstructApproach=Verbalized Distrib.2025.11 | 17.2 | 0.564 | 0.395 | 0.472 | |
| Qwen3-4B-InstructApproach=Verbalized Top-k2025.11 | 16.9 | 0.518 | 0.733 | 0.768 | |
| Qwen3-4B-InstructApproach=Verbalized Conf.2025.11 | 16 | 0.548 | 0.789 | 0.808 |