Benchmark RadarMedGym
Provides a continuous-time reinforcement learning environment for dynamic medical treatment recommendation, built from clinical data with physics-informed neural networks. Supports offline and online RL with configurable parameters.
- Released
- 2026-05-31
- Readiness
- Paper only
- Primary field
- Health & Life Sciences
Why it matters
Offers a realistic testbed for RL methods in continuous-time medical settings, enabling evaluation of personalization, safety, and online deployment gaps. Standardizes comparisons between discrete and continuous time approaches.
Motivation
Medical treatment recommendation poses several challenges to reinforcement learning (RL): patient physiology evolves in continuous time, measurements and interventions are performed at irregular intervals, and treatment effects vary substantially across individuals.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.