Toward a Theory of Hierarchical Memory for Language Agents

TL;DR

This paper introduces a formal framework for hierarchical memory in language agents, unifying various approaches through three core operators and analyzing their interactions and constraints.

Contribution

It proposes a unifying theory with three operators—extraction, coarsening, and traversal—to compare and analyze hierarchical memory systems in language agents.

Findings

Identifies a self-sufficiency spectrum for representative functions

Shows how coarsening and traversal strategies are coupled

Demonstrates the framework's applicability to eleven existing systems

Abstract

Many recent long-context and agentic systems address context-length limitations by adding hierarchical memory: they extract atomic units from raw data, build multi-level representatives by grouping and compression, and traverse this structure to retrieve content under a token budget. Despite recurring implementations, there is no shared formalism for comparing design choices. We propose a unifying theory in terms of three operators. Extraction ($\alpha$) maps raw data to atomic information units; coarsening ($C = (\pi, \rho)$) partitions units and assigns a representative to each group; and traversal ($\tau$) selects which units to include in context given a query and budget. We identify a self-sufficiency spectrum for the representative function $\rho$ and show how it constrains viable retrieval strategies (a coarsening-traversal coupling). Finally, we instantiate the decomposition on eleven existing systems spanning document hierarchies, conversational memory, and agent execution traces, showcasing its generality.