Table 7 Race composition across datasets (count and percentage)
From: Mitigating the risk of health inequity exacerbated by large language models
Dataset | Not mentioned | White | African-Am. | Hispanic | Total |
|---|---|---|---|---|---|
| Â | Count (%) | Count (%) | Count (%) | Count (%) | Â |
MedQA | 1000 (72.1%) | 185 (13.3%) | 90 (6.5%) | 50 (3.6%) | 1387 |
MedMCQA | 6065 (98.6%) | 8 (0.1%) | 1 (0.02%) | 75 (1.2%) | 6149 |
SIGIR 2016 | 40 (69.0%) | 10 (17.2%) | 4 (6.9%) | 2 (3.4%) | 58 |
TREC 2021 | 45 (60.0%) | 15 (20.0%) | 8 (10.7%) | 4 (5.3%) | 75 |
TREC 2022 | 35 (70.0%) | 8 (16.0%) | 3 (6.0%) | 2 (4.0%) | 50 |
