Practice-based clinical mastery through progressively challenging simulated encounters

Design · All designs
Evidence · 3 claims (3 for) · 1 study (1 design), q2 · 0 of 1 report an effect size · 3 claims rest on one study

Description

The article's organizing premise is that clinical competence, for AI as for physicians, is developed through practice rather than acquired through static knowledge testing alone. As the authors state, "ResidencyRL is built on a simple premise: clinical mastery, for AI as for physicians, is developed through practice." The framework draws a direct analogy to medical residency, where trainees convert academic knowledge into clinical expertise through years of supervised encounters with diverse feedback and progressively greater autonomy. It holds that a trainee who scores perfectly on examinations may still fail when presentations are atypical, and that the gap is a failure of clinical process rather than medical knowledge — a process that can be optimized through multi-turn RL in simulation.

Design Implications

Context

Requirements

• Requires simulated environments with sufficient fidelity that competencies trained in simulation transfer to real clinical settings
• Requires a curriculum spanning routine, targeted history-taking, and adversarial scenario types to develop distinct clinical competencies

Constraints

• Prospective validation with real-world workflows remains necessary to establish clinical utility
• The competence ceiling of simulation-based training is bounded by environment fidelity

Target Learners

• Clinical AI agents
• Medical trainees (by analogy to residency education)

Learning Goals

• Sequential clinical reasoning, diagnostic accuracy, management quality, patient-centered communication, and safety-critical decision-making under uncertainty

Claims

• Residencyrl Improves Adversarial Diagnostic Accuracy [+M]
• Residencyrl Generalizes Out Of Domain [+M]
• Residencyrl Reduces Missed Red Flags [+M]

Related Designs

• Five-agent DPT simulated classroom on a static-friction problem

Examples

-

Key Sources

• Liévin, V., Schmidgall, S., Strother, T., Bijamov, A., Goel, A., Palepu, A., Park, C., Balazadeh, V., Sun, M. W., Guerard, M., Chen, J., Steiner, D., Dhillon, V., Azar, I., Mehta, A., Spetsieris, N., Shah, S., Abdelrahim, M., Dahiya, A., Liu, Y., Chou, K., Matias, Y., Hassidim, A., Webster, D. R., Le, Q. V., Hadsell, R., Barral, J., Radebaugh, C., Faust, A., Azizi, S., Schaekermann, M., Chen, P.-H. C., Tu, T., Racz, D., & Yang, L. (2026). ResidencyRL: Reinforcement Learning in Simulated Clinical Environments. arXiv. https://arxiv.org/abs/2608.07418

Page metadata

Field	Value
type	design
id	residencyrl-practice-based-clinical-mastery
title	Practice-based clinical mastery through progressively challenging simulated encounters
description	The article's organizing premise is that clinical competence, for AI as for physicians, is developed through practice rather than acquired through static knowledge testing alone.
status	review
generated	by process:settle-candidates on 2026-10-11
trust	unverified — no one has confirmed this page's content
sources	1 — liévin-2026

Frontmatter is the page's machine-readable record — see the Schema & Guide.