notesum.ai
Published at December 10Zero-Shot ATC Coding with Large Language Models for Clinical Assessments
cs.CL
Released Date: December 10, 2024
Authors: Zijian Chen1, John-Michael Gamble1, Micaela Jantzi2, John P. Hirdes2, Jimmy Lin1
Aff.: 1University of Waterloo; 2University of Waterloo, InterRAI Canada
| Correct Level | Health Canada | RABBITS | Ontario Health | |||||
|---|---|---|---|---|---|---|---|---|
| Fine-tuned 8B* | Llama 3.1 70B | GPT-4o | Fine-tuned 8B* | Llama 3.1 70B | GPT-4o | Fine-tuned 8B* | Llama 3.1 70B | |
| 60.5% | 60.3% | 78.4% | 26.4% | 19.8% | 39.4% | 53.1% | 49.4% | |
| 67.7% | 64.7% | 79.1% | 32.9% | 32.1% | 47.8% | 68.3% | 67.5% | |
| 78.3% | 74.6% | 84.3% | 43.5% | 43.5% | 55.4% | 85.2% | 83.2% | |
| 84.7% | 80.7% | 87.3% | 46.7% | 52.7% | 64.4% | 88.3% | 88.0% | |
| 90.3% | 87.1% | 90.3% | 62.8% | 71.2% | 81.8% | 91.2% | 89.8% | |