Post Persona Alignment for Multi-Session Dialogue Generation

TL;DR

Post Persona Alignment (PPA) tackles long-term persona consistency in multi-session dialogues by reversing the traditional retrieval-grounding flow. It first generates a general response $R_g$ from context $D$, then retrieves relevant memories $M_k$ using $R_g$ as query with threshold $\theta$, and finally refines to a persona-consistent response $R$, aided by a memory pool of structured $(name, relation, object)$ triples. History is compressed into explicit memory to preserve stable persona attributes and improve retrieval precision. Empirical results on multi-session dialogues generated with GPT-3.5 show that PPA improves consistency (C-Score), diversity (Entropy), and personalization (P-F1) over baselines, demonstrating greater naturalness and controllability for long-term personalized dialogue systems.

Abstract

Multi-session persona-based dialogue generation presents challenges in maintaining long-term consistency and generating diverse, personalized responses. While large language models (LLMs) excel in single-session dialogues, they struggle to preserve persona fidelity and conversational coherence across extended interactions. Existing methods typically retrieve persona information before response generation, which can constrain diversity and result in generic outputs. We propose Post Persona Alignment (PPA), a novel two-stage framework that reverses this process. PPA first generates a general response based solely on dialogue context, then retrieves relevant persona memories using the response as a query, and finally refines the response to align with the speaker's persona. This post-hoc alignment strategy promotes naturalness and diversity while preserving consistency and personalization. Experiments on multi-session LLM-generated dialogue data demonstrate that PPA significantly outperforms prior approaches in consistency, diversity, and persona relevance, offering a more flexible and effective paradigm for long-term personalized dialogue generation.

Figures (3)

• Figure 1: Overview of our PPA framework.
• Figure 2: Prompt used for response generation. Curly braces {} indicate variable substitution.
• Figure 3: Prompt used for response refinement.