HyPE: Category-Aware Hypergraph Encoding with Persistent Edge Embeddings for Persona-Grounded Dialogue

HyPE (Hypergraph Persona Encoder) introduces a novel framework for persona-grounded dialogue systems that organizes persona attributes into a hypergraph structure, enhancing the modeling of high-order relations among these attributes. It utilizes an HyperGCN hypergraph neural network to generate a persona summary vector and a soft-memory bank, incorporating Persistent Edge Embeddings (PEE) for improved message passing. This approach outperforms traditional sentence-level pooling methods on the PersonaChat benchmark across various models, including GPT-2, LLaMA-3.2-3B, and Qwen2.5-3B, demonstrating the benefits of structured hyperedge-level encoding for dialogue generation.