Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Context Cartography: Toward Structured Governance of Contextual Space in Large Language Model Systems

Zihua Wu, Georg Gartner

Abstract

The prevailing approach to improving large language model (LLM) reasoning has centered on expanding context windows, implicitly assuming that more tokens yield better performance. However, empirical evidence - including the "lost in the middle" effect and long-distance relational degradation - demonstrates that contextual space exhibits structural gradients, salience asymmetries, and entropy accumulation under transformer architectures. We introduce Context Cartography, a formal framework for the deliberate governance of contextual space. We define a tripartite zonal model partitioning the informational universe into black fog (unobserved), gray fog (stored memory), and the visible field (active reasoning surface), and formalize seven cartographic operators - reconnaissance, selection, simplification, aggregation, projection, displacement, and layering - as transformations governing information transitions between and within zones. The operators are derived from a systematic coverage analysis of all non-trivial zone transformations and are organized by transformation type (what the operator does) and zone scope (where it applies). We ground the framework in the salience geometry of transformer attention, characterizing cartographic operators as necessary compensations for linear prefix memory, append-only state, and entropy accumulation under expanding context. An analysis of four contemporary systems (Claude Code, Letta, MemOS, and OpenViking) provides interpretive evidence that these operators are converging independently across the industry. We derive testable predictions from the framework - including operator-specific ablation hypotheses - and propose a diagnostic benchmark for empirical validation.

Context Cartography: Toward Structured Governance of Contextual Space in Large Language Model Systems

Abstract

The prevailing approach to improving large language model (LLM) reasoning has centered on expanding context windows, implicitly assuming that more tokens yield better performance. However, empirical evidence - including the "lost in the middle" effect and long-distance relational degradation - demonstrates that contextual space exhibits structural gradients, salience asymmetries, and entropy accumulation under transformer architectures. We introduce Context Cartography, a formal framework for the deliberate governance of contextual space. We define a tripartite zonal model partitioning the informational universe into black fog (unobserved), gray fog (stored memory), and the visible field (active reasoning surface), and formalize seven cartographic operators - reconnaissance, selection, simplification, aggregation, projection, displacement, and layering - as transformations governing information transitions between and within zones. The operators are derived from a systematic coverage analysis of all non-trivial zone transformations and are organized by transformation type (what the operator does) and zone scope (where it applies). We ground the framework in the salience geometry of transformer attention, characterizing cartographic operators as necessary compensations for linear prefix memory, append-only state, and entropy accumulation under expanding context. An analysis of four contemporary systems (Claude Code, Letta, MemOS, and OpenViking) provides interpretive evidence that these operators are converging independently across the industry. We derive testable predictions from the framework - including operator-specific ablation hypotheses - and propose a diagnostic benchmark for empirical validation.
Paper Structure (59 sections, 2 equations, 2 figures, 5 tables)

This paper contains 59 sections, 2 equations, 2 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Long-Context Language Models.
  4. Prompt Engineering and Chain-of-Thought.
  5. Retrieval-Augmented Generation.
  6. Memory-Augmented Agent Architectures.
  7. Context Engineering.
  8. Cartographic Generalization and Ubiquitous Cartography.
  9. The Context Cartography Framework
  10. Contextual Space and Epistemic Zones
  11. Black Fog ($\mathcal{B}$): Hallucination.
  12. Gray Fog ($\mathcal{G}$): Drift and Bloat.
  13. Visible Field ($\mathcal{V}$): Overload and Dilution.
  14. Transition-Level Failure Modes.
  15. Salience Geometry of the Visible Field
  16. ...and 44 more sections

Figures (2)

  • Figure 1: The tripartite zonal architecture. Solid arrows show the four named transitions; the dashed arc shows the degenerate $\mathcal{B} \to \mathcal{V}$ path that Context Cartography recommends mediating through $\mathcal{G}$ or operator guards ($\phi$, $\pi^{+}$). Zone-level failure modes appear below each zone.
  • Figure 2: Cartographic pipelines. The inbound pipeline $\lambda \circ \delta \circ \phi \circ \pi^{+} \circ \sigma$ transforms content from $\mathcal{G}$ to $\mathcal{V}$; the outbound pipeline $\pi^{-} \circ \sigma$ archives from $\mathcal{V}$ to $\mathcal{G}$; the dashed loop represents asynchronous maintenance within $\mathcal{G}$.

Theorems & Definitions (11)

  • Definition 1: Contextual Universe
  • Definition 2: Zone Transitions
  • Definition 3: Salience Function
  • Definition 4: Reconnaissance
  • Definition 5: Selection
  • Definition 6: Simplification
  • Definition 7: Aggregation
  • Definition 8: Projection
  • Definition 9: Displacement
  • Definition 10: Layering
  • ...and 1 more