Get Started
Home
Topics
Search
Library
Research questionHow can AI-agent memory survive model upgrades without losing retrieval or repairability?A new model may interpret old notes differently or query existing embeddings differently, even when the memory store is unchanged. Summarization can discard evidence, while embedding changes can make stored information harder to retrieve or repair.
AI Agents
AI Memory
Information Retrieval
Retrieval-Augmented Generation
Latest papersRecent research connected to this question, newest first.Does Your Agent's Memory Survive a Model Upgrade? A Controlled Study of Memory PortabilityEvidence is limited to 48 synthetic histories with randomized answer codes, exact scoring, and two open-weight models below 10 billion parameters. The study compares long-context raw histories, chunked RAG, model-compressed notes, and fixed-schema knowledge graphs under writer swaps, mixed versus full embedding migration, and store-only repair. Its findings are bounded to these configurations and tested migration directions; raw-history repair succeeded in 34 of 48 cases for one tested direction.research paper · Sep 4, 2026
Related questions
How can AI memory remain inspectable and directly correctable during long-term interaction?How can memory-augmented LLM agents coordinate memory construction, retrieval, and repair over long-horizon interactions?How can personalized LLM agents retrieve time-valid memories of persistent and evolving user states?How can research agents reuse operational know-how when repositories are too large to load during tasks?