AI RESEARCH
MemTrace: Tracing and Attributing Errors in Large Language Model Memory Systems
arXiv CS.AI
•
ArXi:2605.28732v1 Announce Type: cross Memory is essential for enabling large language models to long-horizon reasoning, yet existing memory systems remain unreliable and difficult to debug. Tracing memory's dynamic evolution is crucial to understand how information is synthesized, propagated, or corrupted over time. In this work, we study the new problem of error tracing and attribution in LLM memory systems. We propose a novel framework that transforms memory pipelines into executable memory evolution graphs, enabling fine-grained tracing of operational information flow.