How it works

The paper surveys openly available large models adapted via fine-tuning for tasks like diagnosis and record summarization, then maps barriers to real-world use.