Reproduce “DEMENTIA-PLAN: An Agent-Based Framework for Multi-Knowledge Graph Retrieval-Augmented Generation in Dementia Care” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yutong Song, Chenhan Lyu, Pengfei Zhang, Sabine Brunswicker, Nikil Dutt, Amir Rahmani
An agent-based framework named DEMENTIA-PLAN was designed at the University of California Irvine to enhance conversational support for mild-stage dementia patients by integrating multi-knowledge graphs with a self-reflection planning agent. This system demonstrated near human-level performance in empathy and emotional safety, achieving scores of 8.48 and 9.44 respectively, and improved problem-solving by 0.59 points over baseline models.
Key insights
- Combining multi-dimensional knowledge from both daily routine and life memory graphs is crucial for providing both factual accuracy and emotional resonance in dementia care interactions.
- An adaptive, self-reflective planning agent can dynamically optimize knowledge retrieval and integration from heterogeneous sources, significantly improving response coherence and problem-solving over static RAG approaches.
- Incorporating personal memories through AI-assisted support demonstrably enhances emotional stability, empathy, and memory recall for dementia patients, aligning with clinical observations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.