| Task Name | Dataset Name | SOTA Result | Trend | |
|---|---|---|---|---|
| Jailbreaking | EHRAgent TREQS | SR71.78 | 30 | |
| Jailbreaking | EHRAgent eICU | Success Rate (SR)57.93 | 30 | |
| Jailbreaking | EHRAgent MIMIC-III | SR56.55 | 30 | |
| Jailbreaking | EHRAgent ALL | Weighted Average ASR70.112 | 24 | |
| Healthcare Record Management | EHRAgent | Accuracy (ACC)74.8 | 24 | |
| Data Extraction Attack | EHRAgent | Equality (EQ)83 | 20 | |
| Data Extraction Attack on Agent Memory | EhrAgent (test) | Equality (EQ)82 | 12 | |
| Internal memory extraction attack detection | EHRAgent | AUROC1 | 12 |