In the emerging Era of Agents, the core parts of AI systems are being reconstructed to meet the needs not of ordinary LLMs, but of AI agents. We explained what the concept of agentic memory is and why it is becoming a new retrieval layer for agents’ knowledge in our recent article. But here we’ve collected resources that will help you explore this topic much deeper:
Rethinking Memory Mechanisms of Foundation Agents in the Second Half: A Survey
This survey from numerous researchers at Stanford, Salesforce, Google, Meta, and other leading AI organizations explains how AI agents use memory to handle long, complex interactions and real-world tasks, categorizing different memory types, architectures, and evaluation methods, while discussing current challenges and future directions. → Read more
A Survey on the Security of Long-Term Memory in LLM Agents: Toward Mnemonic Sovereignty
Explores the security risks of long-term memory in AI agents – how persistent memory can be manipulated, leaked, corrupted, or spread across systems. The paper organizes threats across the memory lifecycle – writing, storing, retrieving, sharing and deleting information – and argues that future AI systems will need stronger memory governance. → Read more
Graph-based Agent Memory: Taxonomy, Techniques, and Applications
Examines how graphs are used to represent entities and relationships, organize hierarchical knowledge, improve retrieval and reasoning, and continuously update long-term memory for self-evolving AI agents. → Read more
Anatomy of Agentic Memory: Taxonomy and Empirical Analysis of Evaluation and System Limitations
Focuses on practical agentic memory systems, examining memory architectures, benchmark limitations, evaluation reliability, backbone-model dependence, latency and maintenance costs, and why current systems often fail to match their theoretical promise. → Read moreThe AI Hippocampus: How Far are We From Human Memory?
It spans implicit memory in model weights, explicit external memory, and agentic persistent memory, and extends the discussion into multimodal settings where memory must work across vision, language, audio, and action. → Explore more
Why AI Intelligence is Nothing Without Visual Memory | Shawn Shen on the Future of Embodied AI
Shawn Shen argues AI needs a separate, hippocampus-like memory to move beyond chatbots, enabling long-term visual memory, object permanence, and on-device intelligence for robots, wearables, and the physical world. → Watch moreWhen Will We Give AI True Memory? A conversation with Edo Liberty, CEO and founder @ Pinecone
Edo Liberty discusses what real memory in LLMs requires beyond RAG - from scalable vector storage to reliable knowledge systems – and why storage, not compute, is becoming the key bottleneck for building dependable AI agents. → Watch moreDoes AI Remember? The Role of Memory in Agentic Workflows
This is our article about SOAR’s legacy, memory types, generative workflows, memory mode in LLM, and AI influence on memory itself. → Read more
MemEvolve: Meta-Evolution of Agent Memory Systems
MemEvolve is a framework that shows how to jointly adapt agent experience and memory architecture. It also presents EvolveLab, a modular codebase for comparing memory designs, showing improved performance and transfer across tasks and LLMs. → Read moreMemory in the Age of AI Agents
Another great survey that organizes agent memory research. It gives concrete taxonomies across memory form, function, and dynamics, summarizes benchmarks, frameworks, and directions for building systematic agent memory systems. → Read more
