# M2A: Multimodal Memory Agent with Dual-Layer Hybrid Memory for Long-Term Personalized Interactions Source: [M2A: Multimodal Memory Agent with Dual-Layer Hybrid Memory for Long-Term Personalized Interactions]() Collection index: [llms.txt]() Data updated: 2026-02-11T05:00:04.179Z Authors: Junyu Feng, Binxiao Xu, Jiayi Chen, Mengyu Dai, Cenyang Wu, Haodong Li, Bohan Zeng, Yunliu Xie, Hao Liang, Ming Lu, Wentao Zhang Published: 2026-02-07T17:13:56.000Z ## Abstract This work addresses the challenge of personalized question answering in long-term human-machine interactions: when conversational history spans weeks or months and exceeds the context window, existing personalization mechanisms struggle to continuously absorb and leverage users' incremental concepts, aliases, and preferences. Current personalized multimodal models are predominantly static-concepts are fixed at initialization and cannot evolve during interactions. We propose M2A, an agentic dual-layer hybrid memory system that maintains personalized multimodal information through online updates. The system employs two collaborative agents: ChatAgent manages user interactions and autonomously decides when to query or update memory, while MemoryManager breaks down memory requests from ChatAgent into detailed operations on the dual-layer memory bank, which couples a RawMessageStore (immutable conversation log) with a SemanticMemoryStore (high-level observations), providing memories at different granularities. In addition, we develop a reusable data synthesis pipeline that injects concept-grounded sessions from Yo'LLaVA and MC-LLaVA into LoCoMo long conversations while preserving temporal coherence. Experiments show that M2A significantly outperforms baselines, demonstrating that transforming personalization from one-shot configuration to a co-evolving memory mechanism provides a viable path for high-quality individualized responses in long-term multimodal interactions. The code is available at https://github.com/Little-Fridge/M2A. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Multimodal Personalized Question Answering on LoCoMo (test)](): Multimodal Personalized Question Answering; LoCoMo (test); Multi Hop (Q): 48.58. - [Task success rate on DMV-Bench J=5 1.0](): Task success rate; DMV-Bench J=5 1.0; Task Success Rate: 70.4. - [Task success rate on DMV-Bench J=10 1.0](): Task success rate; DMV-Bench J=10 1.0; Task Success Rate: 64.8. - [Task success rate on DMV-Bench J=15 1.0](): Task success rate; DMV-Bench J=15 1.0; Task Success Rate: 62.6. - [Task success rate on DMV-Bench MC J=50 1.0](): Task success rate; DMV-Bench MC J=50 1.0; Task Success Rate: 64.7. ## Original sources and implementation [Original paper]() [PDF]()