Reproduce “MemEmo: Evaluating Emotion in Memory Systems of Agents” 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
Peng Liu, Zhen Tao, Jihao Zhao, Ding Chen, Yansong Zhang, Cuiping Li, Zhiyu Li, Hong Chen
Researchers from Renmin University of China introduce MemEmo, a benchmark framework along with the Human-Like Memory Emotion (HLME) dataset, designed to evaluate the emotional processing capabilities of AI agent memory systems in long-term human-computer interactions. The evaluation demonstrates that current memory systems show varied strengths and weaknesses across emotional information extraction, memory updating, and question answering, indicating no single system robustly handles all emotional aspects.
Key insights
- No single evaluated memory system achieves comprehensive superiority across all emotional processing dimensions, highlighting trade-offs in current architectural designs for emotional intelligence.
- Memory systems like Letta excel in dynamic emotional state updating and robustness in long-context scenarios, while others like Mirix show strengths in short-to-mid-term static emotional extraction.
- Effective emotional memory management requires a critical balance between high-sensitivity emotional perception and stable long- and short-term memory interaction, an area where current systems exhibit limitations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.