Wu, Zhou, Ma, Chen, Gao, Lin, Wu, Gou, Liu, Lau & Dou (2026) — CUHK; Southern Medical University; Peking University; Tencent. arXiv preprint (cs.CL). 📄 Full text (arXiv)
MedGame transforms static clinical cases into structured, executable storytelling games for medical education, moving beyond the localized question-answering and single-turn feedback that characterize most llm medical-training systems. It uses a dual-engine design: a Medical Narrative Designer synthesizes case-grounded clinical storylines with states and decision nodes, while a Story Director converts them into dependency-aware multimodal orchestration plans rendered by an interactive platform. The authors release MedGame Bench, a 5,000-case benchmark and evaluation protocol for Medical Narrative Generation and Story Direction; task-specific fine-tuning substantially improves open-source LLMs and narrows the gap with commercial models. A pilot student study finds learners perceive MedGame as more engaging and useful than text-only alternatives, extending decision-centered, immersive approaches seen in multimodal-ai-tutoring and intelligent-tutoring. Its use of generative-ai for scenario authoring connects to llm-tts-dialogue-lesson-generation, and the engagement gains speak to engagement-metrics as a design target in professional training.
Related Pages
- intelligent-tutoring — decision-centered learning trajectories
- multimodal-ai-tutoring — multimodal orchestration of learning content
- engagement-metrics — learner-perceived engagement outcomes
- generative-ai — LLM-driven scenario and narrative generation
- llm-tts-dialogue-lesson-generation — related LLM-authored interactive lesson content
- genai-patient-education-transplant-handbooks — Grounding genAI patient-education assistants in institution-authored handbooks produces inconsistent