Evidence-Grounded Diagnostic Reasoning on CXReasonDial
99.7CoverageCXReasonAgent
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| CXReasonAgentBackbone=Qwen2.5-7B, GT Setting=With GT (w/)2026.02 | 99.7 | 89.5 | 90 | 95.3 | 4.6 | 80.6 | 23.1 | |
| CXReasonAgentBackbone=Qwen2.5-32B, GT Setting=Without GT (w/o)2026.02 | 99.6 | 98.6 | 97 | 97.6 | 2.3 | 94.2 | 61.1 | |
| CXReasonAgentBackbone=Qwen2.5-7B, GT Setting=Without GT (w/o)2026.02 | 99.6 | 86.3 | 86.6 | 93.7 | 6.3 | 76.6 | 21.1 | |
| CXReasonAgentBackbone=Qwen2.5-1.5B, GT Setting=Without GT (w/o)2026.02 | 99.6 | 91.6 | 92.1 | 94.7 | 5.3 | 80.4 | 31.6 | |
| CXReasonAgentBackbone=Gemini-1.5-Flash, GT Setting=Without GT (w/o)2026.02 | 99.5 | 99.8 | 97.6 | 99.7 | 0.3 | 96.8 | 73.4 | |
| CXReasonAgentBackbone=GPT-4o mini, GT Setting=Without GT (w/o)2026.02 | 99.5 | 99.9 | 98.2 | 99.2 | 0.8 | 96.9 | 74.8 | |
| CXReasonAgentBackbone=GPT-4o mini, GT Setting=With GT (w/)2026.02 | 99.5 | 99.9 | 98.2 | 99.2 | 0.8 | 96.9 | 74.2 | |
| CXReasonAgentBackbone=Qwen2.5-32B, GT Setting=With GT (w/)2026.02 | 99.5 | 97 | 97.3 | 98.6 | 1.4 | 91.6 | 59 | |
| CXReasonAgentBackbone=Gemini-1.5-Flash, GT Setting=With GT (w/)2026.02 | 99.4 | 99.8 | 97.6 | 99.7 | 0.3 | 96.6 | 72.2 | |
| CXReasonAgentBackbone=Llama 3.1-70B, GT Setting=Without GT (w/o)2026.02 | 99.4 | 99.3 | 96.7 | 99.3 | 0.7 | 95.3 | 65.8 | |
| CXReasonAgentBackbone=Llama 3.1-70B, GT Setting=With GT (w/)2026.02 | 99.4 | 99.4 | 96.9 | 99.2 | 0.8 | 95.5 | 65.2 | |
| CXReasonAgentBackbone=Qwen2.5-1.5B, GT Setting=With GT (w/)2026.02 | 99.4 | 92.6 | 93.7 | 96.3 | 3.7 | 82.2 | 32.9 | |
| MedGemma 27BGT Setting=Without GT (w/o)2026.02 | 98.9 | — | — | 53.9 | 45.4 | 46.1 | 6.1 | |
| Pixtral-LargeGT Setting=Without GT (w/o)2026.02 | 98.8 | — | — | 57.9 | 41.5 | 48.6 | 5.9 | |
| Pixtral-LargeGT Setting=With GT (w/)2026.02 | 98.8 | — | — | 79.8 | 20.1 | 65.4 | 6 | |
| MedGemma 27BGT Setting=With GT (w/)2026.02 | 98.5 | — | — | 76.2 | 23.2 | 64.1 | 6.5 | |
| Gemini-1.5-FlashCategory=LVLM Baseline, GT Setting=Without GT (w/o)2026.02 | 98.3 | — | — | 43.1 | 55.6 | 36.9 | 8.6 | |
| Gemini-1.5-FlashCategory=LVLM Baseline, GT Setting=With GT (w/)2026.02 | 98.3 | — | — | 81.1 | 18 | 69.1 | 9.8 |