Research and validation

Trust should be earned with methods, not adjectives.

Lymph evaluates the evidence pipeline in layers and distinguishes technical testing from clinical validation.

What we examine

Evaluation follows the path from retrieval to rendered answer.

Whether relevant supporting evidence can be found across intended source classes.

Whether an answer's important claims are traceable to sources that actually support them.

Whether high-risk cases, uncertainty, contraindications, and abstention behavior are handled appropriately.

Whether operations complete, stream, resume, and preserve their results consistently.

Publication standard

A dated and versioned evaluation set

A documented scoring method

Coverage and abstention reported separately

Known limitations and contamination checks

A reproducible result and responsible owner

Automated regression tests are not prospective clinical validation.
A citation marker does not prove claim-level faithfulness.
Performance on one specialty or question family does not establish performance across medicine.
Lymph is not presented as autonomous diagnosis or prescribing software.
No public accuracy, adoption, time-saving, or compliance headline metric is asserted on this site.

For clinical and governance teams

Hospitals and clinical collaborators can discuss a bounded pilot with explicit methods and review criteria.