Automatic Diagnosis on DxBench
60.1Accuracy (w/o inquiry)GPT-4
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GPT-4Learning Setting=Zero-shot Setting2024.07 | 60.1 | 62.5 | 0.6 | |
| Claude-3-OpusLearning Setting=Zero-shot Setting2024.07 | 59.4 | 62.1 | 1.8 | |
| ERNIE BotLearning Setting=Zero-shot Setting2024.07 | 58.4 | 59.9 | 0.4 | |
| DiagnosisGPTLearning Setting=Zero-shot Setting, Confidence threshold (τ)=0.42024.07 | 56.9 | 61.3 | 0.7 | |
| DiagnosisGPTLearning Setting=Zero-shot Setting, Confidence threshold (τ)=0.52024.07 | 56.9 | 62.4 | 1.8 | |
| DiagnosisGPTLearning Setting=Zero-shot Setting, Confidence threshold (τ)=0.62024.07 | 56.9 | 64.5 | 3.1 | |
| Gemini-ProLearning Setting=Zero-shot Setting2024.07 | 55.3 | 55.1 | 0.1 | |
| Yi-34B-ChatLearning Setting=Zero-shot Setting2024.07 | 52.4 | 54.6 | 0.2 | |
| GPT-3.5Learning Setting=Zero-shot Setting2024.07 | 52.1 | 52.3 | 0.1 | |
| Mixtral-8x7B-Instruct-v0.1Learning Setting=Zero-shot Setting2024.07 | 42.1 | 41.2 | 1.4 |