Sign InOpen Brain
arXivPaperNeeds Review

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

MDL adds a white-box trust decision between retrieval and generation, using relevance, reliability, and task risk to reject conflicting memories without another model call.

arXiv · Sep 18, 2026
Open Source Open MarkdownOpen JSON
Source Summary

The Memory Decision Layer is a **zero-parameter** controller between retrieval and generation. It combines relevance, reliability, and task risk, separates confidence from consistency, and can abstain instead of injecting a retrieved memory.

Practical Implication

Agent builders should treat retrieval and trust as separate decisions, especially when stored memories conflict. In the reported evaluations, MDL reduced hallucinations by **about 56.04%** in general conflict scenarios and added **about 0.14 ms** per decision.

Agent-Ready Context
The Memory Decision Layer is a **zero-parameter** controller between retrieval and generation. It combines relevance, reliability, and task risk, separates confidence from consistency, and can abstain instead of injecting a retrieved memory.

Agent builders should treat retrieval and trust as separate decisions, especially when stored memories conflict. In the reported evaluations, MDL reduced hallucinations by **about 56.04%** in general conflict scenarios and added **about 0.14 ms** per decision.

The results span multiple models and open datasets, but the abstract does not identify them or describe production workloads. Near-zero hallucination in high-risk scenarios therefore remains a paper result rather than a deployment guarantee.
Connected Context · Feed7 Judgment

MDL makes memory admission a distinct, inspectable control after retrieval: relevance alone does not justify injecting a memory when reliability is low or task risk is high. It complements query-scoped suppression and provenance-aware stores with a cheap abstention point, but does not address model-upgrade portability or establish that its reported conflict results transfer to production workloads.

MeClear: Cooperative Game-Theoretic Attribution and Risk-Aware Memory Clearance for Long-Horizon LLM AgentsMeClear suppresses harmful combinations after retrieval, while MDL decides whether retrieved memory should be injected from relevance, reliability, and risk; together they cover complementary conflict controls.Citation Needed: Provenance for LLM-Built Knowledge Graphs — Daniel Chalef, Zep AIProvenance supplies evidence that a reliability signal could use, especially when synthesized memories merge, change, or outlive their sources.Does Your Agent's Memory Survive a Model Upgrade? A Controlled Study of Memory PortabilityMemory-portability failures remain upstream of MDL: unchanged decision rules cannot ensure reliable admission if a model upgrade or mixed embedding index changes what gets retrieved.Your agents lack context: Here's how to fix "You're absolutely right!" — Brandon Waselnuk, UnblockedThe context-engine proposal likewise separates retrieval from conflict ranking and permissions; MDL adds an explicit risk-aware abstention mechanism to that design.
Context Map
context#agent-memory#retrieval#agent-reliability
Uncertainty
The results span multiple models and open datasets, but the abstract does not identify them or describe production workloads. Near-zero hallucination in high-risk scenarios therefore remains a paper result rather than a deployment guarantee.