🤖 AI Summary
To address the degradation of retrieval performance in RAG-based chatbots during prolonged dialogues—caused by memory bloat—this paper proposes LUFY, a lightweight, emotion-aware memory filtering method inspired by cognitive psychology. LUFY models the importance of dialogue memories and selectively retains only <10% of high-value segments identified by high emotional arousal, enabling proactive forgetting. It represents the first systematic integration of “emotion-enhanced memory” and “selective forgetting” mechanisms into RAG long-term memory management. We introduce the longest (8 hours per user), multi-turn, real-world interactive evaluation paradigm to date. Experiments demonstrate that LUFY significantly improves long-horizon retrieval accuracy and user satisfaction, sustaining stable performance over continuous 8-hour dialogues—empirically validating that selective forgetting outperforms exhaustive memory retention.
📝 Abstract
While Retrieval-Augmented Generation (RAG) has shown promise in enhancing long-term conversations, the increasing memory load as conversations progress degrades retrieval accuracy. Drawing on psychological insights, we propose LUFY, a simple yet effective method that focuses on emotionally arousing memories and retains less than 10% of the conversation. In the user experiment, participants interacted with three types of RAG chatbots, each for 2 hours over 4 sessions, marking the most extensive assessment of a chatbot's long-term capabilities to date -- more than four times longer than any existing benchmark. The results demonstrate that prioritizing arousing memories while forgetting the majority of the conversation significantly enhances user experience. This study pushes the frontier of long-term conversations and highlights the importance of forgetting unimportant parts of conversations. Code and Dataset: https://github.com/ryuichi-sumida/LUFY