Large Language Models Underperform in Real-World ED Triage Benchmark
Benchmarking 16 large language model configurations on 16,107 emergency triage encounters found mixed accuracy, weak sector assignment, and variability.
This section contains the website’s posts.