What it measures

Simulated patient conversations test whether a model gathers a useful history and reaches the correct diagnosis. Compare conversational results with vignette questions and keep free-response answers separate from multiple choice.

Reading the measurement

Accuracy is only part of the picture. Inspect unsupported answers, abstentions, language coverage and the evidence available to the model. A confident response can still be wrong, and a broad average can hide a specialist gap.

Published evaluation

Diagnostic accuracy on the combined 2,000-case evaluation. Free-response and multiple-choice answer formats, and four ways of presenting case information, remain separate. The live board ranks free-response multi-turn conversation performance. These are simulated clinical interactions, not evidence of patient outcomes.

Where this fits in an evaluation suite

Read accuracy together with error and abstention rates. Domain averages can hide language gaps, specialist failures, and poorly calibrated confidence.

Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.

A reproducible comparison

Diagnostic accuracy on the combined 2,000-case evaluation. Free-response and multiple-choice answer formats, and four ways of presenting case information, remain separate. The live board ranks free-response multi-turn conversation performance. These are simulated clinical interactions, not evidence of patient outcomes.

Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.

PRIMARY EVIDENCE & FURTHER READING

  1. An evaluation framework for clinical use of large language models in patient interaction tasks — official CRAFT-MD leaderboard ↗

    Shreya Johri, Jaehwan Jeong, Benjamin A. Tran, Daniel I. Schlessinger, Shannon Wongvibulsin, Leandra A. Barnes et al. · 2025

    All 12 authors

    Shreya Johri, Jaehwan Jeong, Benjamin A. Tran, Daniel I. Schlessinger, Shannon Wongvibulsin, Leandra A. Barnes, Hong-Yu Zhou, Zhuo Ran Cai, Eliezer M. Van Allen, David Kim, Roxana Daneshjou, Pranav Rajpurkar

Reviewed 2026-10-27. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.