{
  "schema_version": "1.1",
  "id": "s13:https://arxiv.org/abs/2609.22043v1",
  "slug": "2609-22043v1-0442mc1",
  "url": "https://feed7.dev/p/2609-22043v1-0442mc1",
  "title": "An Interpretable Memory Decision Controller for LLM Agents Based on Three-Signal Complementarity: Decoupling Confidence and Consistency",
  "why_included": "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.",
  "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_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.\n\nAgent 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.\n\nThe 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.",
  "source": {
    "name": "arXiv",
    "url": "https://arxiv.org/abs/2609.22043v1",
    "published_at": "2026-09-18T17:34:13.000Z"
  },
  "source_class": "blog_post",
  "content_type": "Paper",
  "layer": "context",
  "domains": [],
  "topics": [
    "agent-memory",
    "retrieval",
    "agent-reliability"
  ],
  "verification": {
    "status": "needs_review",
    "label": "Needs Review",
    "method": "unverified",
    "verified_at": null
  },
  "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."
  ],
  "connected_context": {
    "meaning": "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.",
    "corpus_size": 831,
    "generated_at": "2026-09-21T09:04:18.040Z",
    "connections": [
      {
        "title": "MeClear: Cooperative Game-Theoretic Attribution and Risk-Aware Memory Clearance for Long-Horizon LLM Agents",
        "source_name": "arXiv",
        "source_url": "https://arxiv.org/abs/2609.09115v1",
        "feed7_url": "https://feed7.dev/p/2609-09115v1-1xq0kqg",
        "reason": "MeClear 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."
      },
      {
        "title": "Citation Needed: Provenance for LLM-Built Knowledge Graphs — Daniel Chalef, Zep AI",
        "source_name": "AI Engineer",
        "source_url": "https://www.youtube.com/watch?v=H7puB0RwJMM",
        "feed7_url": "https://feed7.dev/p/citation-needed-provenance-for-llm-built-knowledge-graphs-daniel-chalef-1iob5t8",
        "reason": "Provenance supplies evidence that a reliability signal could use, especially when synthesized memories merge, change, or outlive their sources."
      },
      {
        "title": "Does Your Agent's Memory Survive a Model Upgrade? A Controlled Study of Memory Portability",
        "source_name": "arXiv",
        "source_url": "https://arxiv.org/abs/2609.05339v1",
        "feed7_url": "https://feed7.dev/p/2609-05339v1-06dws1c",
        "reason": "Memory-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."
      },
      {
        "title": "Your agents lack context: Here's how to fix \"You're absolutely right!\" — Brandon Waselnuk, Unblocked",
        "source_name": "AI Engineer",
        "source_url": "https://www.youtube.com/watch?v=KcVkq5L-0f0",
        "feed7_url": "https://feed7.dev/p/your-agents-lack-context-here-s-how-to-fix-you-re-absolutely-right-brand-0pdoj93",
        "reason": "The context-engine proposal likewise separates retrieval from conflict ranking and permissions; MDL adds an explicit risk-aware abstention mechanism to that design."
      }
    ]
  },
  "lifecycle": "Current",
  "published_at": "2026-09-18T17:34:13.000Z",
  "modified_at": "2026-09-18T17:34:13.000Z",
  "supersedes": [],
  "expires_at": null,
  "formats": {
    "html": "https://feed7.dev/p/2609-22043v1-0442mc1",
    "json": "https://feed7.dev/p/2609-22043v1-0442mc1.json",
    "markdown": "https://feed7.dev/p/2609-22043v1-0442mc1.md"
  }
}