An Interpretable Memory Decision Controller for LLM Agents Based on Three-Signal Complementarity: Decoupling Confidence and Consistency
A zero-parameter Memory Decision Layer cuts LLM hallucination from conflicting retrieved memories by about 56 percent.
The Memory Decision Layer is a zero-parameter controller placed between retrieval and generation that decides whether an LLM agent should trust retrieved memories. Its encoder combines relevance, reliability, and task risk, separates confidence from consistency, and can abstain. On mainstream large language models, it cuts hallucination under conflicting memories by about 56.04 percent in general settings and approaches zero hallucination in high-risk settings. Decisions use only geometric operations and add about 0.14 ms, roughly 50 times faster than the preceding embedding retrieval.
- Conflicting retrieved memories can raise RAG hallucination above a memory-free baseline.
- MDL fuses relevance, reliability, and task risk with no trained parameters.
- Hallucination falls about 56 percent generally and near zero in high-risk cases.
- Each decision adds about 0.14 ms, far faster than an LLM self-check.
Full article226 words · extracted from arxiv.org · click to collapse
Memory systems for large language models have focused predominantly on efficient retrieval, whereas the decision of whether retrieved memories should be trusted has received comparatively little attention. When the memory store contains conflicting positions, standard retrieval-augmented generation (RAG) blindly injects memories and amplifies hallucinations: in models susceptible to memory injection, the RAG hallucination rate under conflicting memories is markedly higher than that of a memory-free baseline. Inspired by memory signaling mechanisms in the prefrontal cortex, we propose the Memory Decision Layer (MDL), a zero-parameter memory decision controller situated between the retrieval and generation stages. Its core is a three-signal complementary encoder that fuses relevance, reliability, and task risk through QR-based orthogonal subspace projection and a meta-working-memory signal into an interpretable decision representation that quantifies the trustworthiness of retrieved memories. Building on this encoder, MDL explicitly decouples confidence from consistency and introduces risk inversion and explicit abstention. Evaluations on mainstream large language models and multiple open-source datasets show that MDL reduces the hallucination rate under conflicting memories by about 56.04% in general scenarios and approaches zero hallucination in high-risk scenarios. The controller is fully white-box: it relies purely on geometric operations, requires no trained parameters, and adds only about 0.14 ms per decision -- roughly 50x faster than the embedding-retrieval step that precedes it and four to five orders of magnitude faster than an LLM self-evaluation call.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.22043