This lunchtime lecture series blends innovation, expertise, and a relaxed atmosphere, offering an exploration of the latest applications and future advancements of artificial intelligence in digital medicine. While enjoying a delicious lunch, participants will gain cutting-edge insights into AI in medicine and engage in in-depth discussions with experts.

Speakers:

• Hung-Yu Kao (Professor, Department of Computer Science, National Tsing Hua University)

This talk explored the advances and challenges of generative AI and Natural Language Processing (NLP) in clinical dialogue, logical reasoning, and personalized generation. Using a hospital-collaborative educational project as an example, it demonstrated how an “AI Intern” system can be developed through fine-tuning and prompting techniques. The system is designed to simulate physician–patient interactions and train healthcare professionals to respond to diverse clinical scenarios.

The presentation also highlighted the issue of hallucinations in domain-specific knowledge within large language models and proposed the IDEAL-RAG framework to reconcile parametric memory with external retrieval, thereby improving response reliability. For emotional support, the EED (Explicit Empathy Decomposition) prompting method was introduced, enabling chatbots to generate highly empathetic responses and helping reduce emotional fatigue among healthcare staff. In addition, the talk examined personalized dialogue generation by leveraging Graph Neural Networks (GNNs) to model discourse relations in conversations, addressing the limitations of conventional systems that often produce uniform and incoherent responses.

Finally, the lecture discussed whether large language models possess reasoning capabilities and introduced decoding strategies. It concluded that factual reasoning ability is latent within pre-trained models and can be elicited and strengthened through iterative fine-tuning, aggregation, and retrieval.

We welcome all colleagues interested in AI and medical innovation to sign up and explore the limitless possibilities of smart medicine together!

【Lecture Information】

•  Topic: From Medical Dialogue Generation to Factual Reasoning in Large Language Models 
•  Date & Time: April 15, 2026 (Wed.) | 12:10-13:20
•  Venue: Room IR630, 6th Floor, International Academic Research Building, Kaohsiung Medical University
•  Registration Deadline: Until April 13, 2026 (Mon.) at 12:00 | Online registration required
•  Registration Link: https://forms.gle/cdmaBN42vUpaLq6R9
•  Participation Mode: Hybrid (In-person: 60 participants; Online: Unlimited)
•  Organizers: Biomedical Artificial Intelligence Academy of KMU, Department of Artificial Intelligence in Medicine of KMU, Medical AI Innovation and Application Center of KMUH
•  Additional Info: Lunch will be provided for in-person attendees. 

※ This lecture is eligible for KMU faculty growth credits. Participants must complete both sign-in/out and the satisfaction survey to receive credit points.

We look forward to your participation!