Procedural Memory Is Not All You Need: Bridging Cognitive Gaps in LLM-Based Agents

TL;DR

The paper addresses the limitations of procedural memory in LLMs for autonomous decision-making within wicked learning environments. It proposes a modular architecture that augments LLMs with semantic and associative memory via independent agentic learners that provide contextual metadata to LLM actors. It analyzes the core constraints of LLMs, outlines a three-part modular design, and details integration through prompt augmentation and cross-module reasoning. The work aims to enable adaptive, memory-rich, and interpretable agents capable of operating effectively in non-stationary real-world settings.

Abstract

Large Language Models (LLMs) represent a landmark achievement in Artificial Intelligence (AI), demonstrating unprecedented proficiency in procedural tasks such as text generation, code completion, and conversational coherence. These capabilities stem from their architecture, which mirrors human procedural memory -- the brain's ability to automate repetitive, pattern-driven tasks through practice. However, as LLMs are increasingly deployed in real-world applications, it becomes impossible to ignore their limitations operating in complex, unpredictable environments. This paper argues that LLMs, while transformative, are fundamentally constrained by their reliance on procedural memory. To create agents capable of navigating ``wicked'' learning environments -- where rules shift, feedback is ambiguous, and novelty is the norm -- we must augment LLMs with semantic memory and associative learning systems. By adopting a modular architecture that decouples these cognitive functions, we can bridge the gap between narrow procedural expertise and the adaptive intelligence required for real-world problem-solving.