HealthCraft: A Reinforcement Learning Safety Environment for Emergency Medicine
Frontier language models are being deployed into clinical workflows faster than the infrastructure to evaluate them safely. Static medical-QA benchmarks miss the failure modes that matter in emergency medicine: trajectory-level safety collapse, tool misuse, and capitulation under sustained clinical pressure. We present HealthCraft, the first public reinforcement-learning environment that rewards trajectory-level safety under realistic emergency-medicine conditions, adapted from Corecraft. It is built on a FHIR R4 world state with 14 entity types and 3,987 seed entities, exposes 24 MCP tools, and defines a dual-layer rubric that zeroes reward whenever any safety-critical criterion is violated. We release 195 tasks across six categories, graded against 2,255 binary criteria (515 safety-critical); a post-hoc 10-task negative-class slate extends this to 205 tasks and 2,337 criteria. V8 results on two frontier models show Claude Opus 4.6 at Pass@1 24.8% [21.5-28.4] and GPT-5.4 at 12.6% [10.2-15.6], with safety-failure rates of 27.5% and 34.0%. On multi-step workflows - the closest proxy to real emergency care - performance collapses to near zero (Claude 1.0%, GPT-5.4 0.0%) despite partial competence on individual steps. Six infrastructure bugs fixed between pilots v2 and v8 re-ordered which model "looks stronger," evidence that infrastructure fidelity is part of the measurement. A deterministic LLM-judge overlay bounds evaluator noise, and a 60-run negative-class smoke pilot shows the reward signal is not drop-in training-safe: restraint criteria pass at 0.929 prevalence, a gameability an eval harness can tolerate but a training reward cannot. We scaffold coupling to a Megatron+SGLang+GRPO loop per Corecraft Section 5.2 and leave training-reward ablations as future work. Environment, tasks, rubrics, and harness are released under Apache 2.0.
Code (0)
등록된 구현이 없습니다.
Tasks
Reinforcement LearningSimilar Papers 제목 키워드 기반
SAFER: Safe Collision Avoidance using Focused and Efficient Trajectory Search with Reinforcement Learning
Collision avoidance is key for mobile robots and agents to operate safely in the real world. In this work we present SAFER, an efficient and effective collision avoidance system that is able to improve safety by correcti…
Collision Avoidancereinforcement-learningReinforcement Learning (RL)Trajectory PlanningSafe Reinforcement Learning for Grid Voltage Control
Under voltage load shedding has been considered as a standard approach to recover the voltage stability of the electric power grid under emergency conditions, yet this scheme usually trips a massive amount of load ineffi…
reinforcement-learningReinforcement LearningReinforcement Learning (RL)Safe Reinforcement LearningIntelligent Emergency Message Broadcasting in VANET Using PSO
The new type of Mobile Ad hoc Network which is called Vehicular Ad hoc Networks (VANET) created a fertile environment for research. In this research, a protocol Particle Swarm Optimization Contention Based Broadcast (PCB…
Barrier Function-based Safe Reinforcement Learning for Emergency Control of Power Systems
Under voltage load shedding has been considered as a standard and effective measure to recover the voltage stability of the electric power grid under emergency and severe conditions. However, this scheme usually trips a …
Deep Reinforcement Learningreinforcement-learningReinforcement Learning (RL)Safe Reinforcement LearningFrom Coordination to Personalization: A Trust-Aware Simulation Framework for Emergency Department Decision Support
Background/Objectives: Efficient task allocation in hospital emergency departments (EDs) is critical for operational efficiency and patient care quality, yet the complexity of staff coordination poses significant challen…
Decision Making