Skip to content

Source paper

Memory in the Age of AI Agents

Yuyang Hu, Shichun Liu, Yanwei Yue and 44 more

Yuyang Hu, Shichun Liu, Yanwei Yue, Guibin Zhang, Boyang Liu, Fangyi Zhu, Jiahang Lin, Honglin Guo, Shihan Dou, Zhiheng Xi, Senjie Jin, Jiejun Tan, Yanbin Yin, Jiongnan Liu, Zeyu Zhang, Zhongxiang Sun, Yutao Zhu, Hao Sun, Boci Peng, Zhenrong Cheng, Xuanbo Fan, Jiaxin Guo, Xinlei Yu, Zhenhong Zhou, Zewen Hu, Jiahao Huo, Junhao Wang, Yuwei Niu, Yu Wang, Zhenfei Yin, Xiaobin Hu, Yue Liao, Qiankun Li, Kun Wang, Wangchunshu Zhou, Yixin Liu, Dawei Cheng, Qi Zhang, Tao Gui, Shirui Pan, Yan Zhang, Philip Torr, Zhicheng Dou, Ji-Rong Wen, Xuanjing Huang, Yu-Gang Jiang, Shuicheng Yan

arXiv 2512.13564 · 2025101 upvotes on Hugging Face

Research ideas from this paper (4)

Generated by AI. The paper’s authors did not write these.

Dynamic Hierarchical Layout Optimization for Token-Level Memory

A reinforcement learning memory architect that dynamically reorganizes hierarchical token graphs by pruning redundant paths and spawning abstraction layers based on real-time retrieval success.

Memory in the Age of AI Agents
🔴Hard🔬Significant ResearchLarge clusterNo close match found#Retrieval & memory#Efficiency

Interpretable Latent Memory via Symbolic Bottleneck Encoders

A hybrid encoder that compresses latent memory vectors through a sparse conceptual bottleneck and generates verified natural language descriptions to ensure interpretable and debuggable model memory.

Memory in the Age of AI Agents
🔴Hard💡Moderate InnovationMultiple GPUsRelated work found#Retrieval & memory#Safety & alignment

Omnimodal Conflict Resolution in Memory Evolution

A cross-modal Bayesian update mechanism that assigns channel reliability weights and detects semantic contradictions across different sensory inputs before updating persistent memory bases.

Memory in the Age of AI Agents
🔴Hard💡Moderate InnovationSingle GPU or lessNo close match found#Retrieval & memory#Multimodal

Self-Rationalizing Post-Retrieval Filtering via Counterfactual Attribution

A counterfactual attribution module that evaluates what-if scenarios by removing individual memory snippets during re-ranking, automatically pruning low-impact fragments and outputting traceable justification logs.

Memory in the Age of AI Agents
🟠Medium💡Moderate InnovationSingle GPU or lessRelated work found#Retrieval & memory#Reasoning

AI summary

This survey provides a comprehensive taxonomy of AI agent memory, categorizing research through the lenses of forms (token-level, parametric, latent), functions (factual, experiential, working), and dynamics (formation, evolution, retrieval). It consolidates the fragmented field of agentic intelligence by delineating agent memory from related concepts like RAG and context engineering while identifying emerging frontiers in automation, reinforcement learning integration, and multimodal systems.

How it works (AI summary)

The paper uses a multi-dimensional analytical framework called the 'Forms-Functions-Dynamics' triangle to organize existing literature and propose a unified vocabulary for the architectural realization, functional purpose, and lifecycle operations of memory in LLM-based agents.