MIRROR: Cognitive inner monologue for persistent reflection and reasoning in conversational LLMs.
A cognitive architecture that gives language models an inner monologue: parallel reflection across goals, reasoning, and memory, synthesized between turns into a persistent self-narrative.
Arcarae · San Francisco, CA
Abstract
Multiple cognitive theories (Global Workspace Theory, reconstructive episodic memory, inner speech, and complementary learning systems) converge on a shared set of architectural principles for reflection.
MIRROR operationalizes each as a concrete mechanism: an Inner Monologue Manager generates parallel cognitive threads across Goals, Reasoning, and Memory, and a Cognitive Controller synthesizes them into a persistent self-narrative between turns. It separates the Talker, which generates immediate responses from the latest narrative, from the Thinker, which processes each turn asynchronously through parallel reflection.
Evaluated on multi-turn dialogue requiring constraint maintenance under attentional interference, MIRROR yields a 21% relative improvement across seven architecturally diverse language models, reducing sycophancy and contextual drift.