Evaluating Structured Output Robustness of Small Language Models for Open Attribute-Value Extraction from Clinical Notes
This provides practical guidance for deploying language models in privacy-sensitive clinical settings, though it is incremental as it compares existing formats.
The paper tackled the problem of structured output robustness for small language models in clinical note extraction, finding that JSON consistently yields the highest parseability, with improvements from targeted prompting and larger models but declines for longer documents and certain note types.
We present a comparative analysis of the parseability of structured outputs generated by small language models for open attribute-value extraction from clinical notes. We evaluate three widely used serialization formats: JSON, YAML, and XML, and find that JSON consistently yields the highest parseability. Structural robustness improves with targeted prompting and larger models, but declines for longer documents and certain note types. Our error analysis identifies recurring format-specific failure patterns. These findings offer practical guidance for selecting serialization formats and designing prompts when deploying language models in privacy-sensitive clinical settings.