D-MEM: Dopamine-Gated Agentic Memory via Reward Prediction Error Routing

TL;DR

D-MEM is a biologically inspired memory architecture for autonomous language model agents that efficiently manages long-term memory updates by using reward prediction error signals, significantly reducing costs and improving reasoning and resilience.

Contribution

Introduces D-MEM, a novel dopamine-gated memory system that decouples interaction from restructuring, enabling scalable, cost-efficient long-term memory management in autonomous agents.

Findings

Reduces token consumption by over 80%.

Eliminates O(N^2) bottlenecks in memory updates.

Outperforms baselines in reasoning and adversarial tests.

Abstract

Autonomous LLM agents require structured long-term memory, yet current "append-and-evolve" systems like A-MEM face O(N^2) write-latency and excessive token costs. We introduce D-MEM (Dopamine-Gated Agentic Memory), a biologically inspired architecture that decouples short-term interaction from cognitive restructuring via a Fast/Slow routing system based on Reward Prediction Error (RPE). A lightweight Critic Router evaluates stimuli for Surprise and Utility. Routine, low-RPE inputs are bypassed or cached in an O(1) fast-access buffer. Conversely, high-RPE inputs, such as factual contradictions or preference shifts, trigger a "dopamine" signal, activating the O(N) memory evolution pipeline to reshape the agent's knowledge graph. To evaluate performance under realistic conditions, we introduce the LoCoMo-Noise benchmark, which injects controlled conversational noise into long-term sessions. Evaluations demonstrate that D-MEM reduces token consumption by over 80%, eliminates O(N^2) bottlenecks, and outperforms baselines in multi-hop reasoning and adversarial resilience. By selectively gating cognitive restructuring, D-MEM provides a scalable, cost-efficient foundation for lifelong agentic memory.