LLM Filtering on Manually adjudicated gold-standard CUIs Fluid Overload v1 (test)
77CUIs CountGPT-5-mini
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| GPT-5-miniFramework=CUICurate, Backbone=GPT-5-mini2026.02 | 77 | 50 | 92 | 65 | |
| GPT-5Framework=CUICurate, Backbone=GPT-52026.02 | 50 | 34 | 96 | 50 | |
| manualMode=manual curation2026.02 | 30 | 18 | 87 | 30 |