An Interpretable Memory Decision Controller for LLM Agents Based on Three-Signal Complementarity: Decoupling Confidence and Consistency

Yiming Zhang, Jinghong Zhang, Haoran Zhao, Yiren Ma, Chunlei Zhao|September 18, 2026arXiv

Key Takeaway

When using RAG with LLMs, blindly trusting all retrieved memories causes hallucinations; a lightweight geometric decision layer can filter unreliable memories without any learned parameters, making RAG safer and more trustworthy.

Summary

This paper introduces Memory Decision Layer (MDL), a parameter-free controller that decides whether to trust retrieved memories in RAG systems. It uses three signals—relevance, reliability, and task risk—combined through geometric operations to detect conflicting memories and prevent hallucinations, reducing errors by 56% when memories contradict each other.

reasoning safety efficiency

Key Terms

rag-pipeline hallucination memory-injection orthogonal-projection meta-working-memory