Social Determinants of Health Event Extraction on n2c2/UW shared task 2022 (test)
93.8PrecisionOur method (gemini-2.5-flash)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Our method (gemini-2.5-flash)LLM Model=gemini-2.5-flash, SDOH Category=Living Status2026.04 | 93.8 | 88.3 | 91 | |
| Our method (o4-mini)LLM Model=o4-mini, SDOH Category=Tobacco2026.04 | 93.3 | 85.4 | 89.2 | |
| Our method (o4-mini)LLM Model=o4-mini, SDOH Category=Alcohol2026.04 | 91.8 | 89.1 | 90.4 | |
| Our method (o4-mini)LLM Model=o4-mini, SDOH Category=Overall2026.04 | 90.2 | 83.3 | 86.6 | |
| MicrosoftTeam=Microsoft2026.04 | 89.1 | 88.7 | 88.9 | |
| Our method (o4-mini)LLM Model=o4-mini, SDOH Category=Living Status2026.04 | 89 | 86.3 | 87.6 | |
| Our method (o4-mini)LLM Model=o4-mini, SDOH Category=Employment2026.04 | 88.9 | 78.6 | 84.4 | |
| Philips Research North AmericaTeam=Philips Research North America2026.04 | 88.5 | 78 | 82.9 | |
| University of Texas at San AntonioTeam=University of Texas at San Antonio2026.04 | 88 | 84.1 | 86 | |
| Our method (o4-mini)LLM Model=o4-mini, SDOH Category=Drug2026.04 | 87.9 | 74.4 | 80.5 | |
| Children’s Hospital of PhiladelphiaTeam=Children’s Hospital of Philadelphia2026.04 | 87.4 | 88.8 | 88.1 | |
| Our method (gemini-2.5-flash)LLM Model=gemini-2.5-flash, SDOH Category=Employment2026.04 | 86.8 | 85.2 | 86 | |
| Our method (gemini-2.5-flash)LLM Model=gemini-2.5-flash, SDOH Category=Tobacco2026.04 | 84.1 | 84.1 | 84.1 | |
| Our method (gemini-2.5-flash)LLM Model=gemini-2.5-flash, SDOH Category=Overall2026.04 | 83.3 | 81.7 | 82.5 | |
| Our method (gemini-2.5-flash)LLM Model=gemini-2.5-flash, SDOH Category=Alcohol2026.04 | 78.2 | 82.3 | 80.2 | |
| Our method (gemini-2.5-flash)LLM Model=gemini-2.5-flash, SDOH Category=Drug2026.04 | 74.4 | 68.8 | 71.5 |