Knowledge Architecture:ConceptsObservationsEvidence

Token Efficiency Ratio (TER)

TER

Information Density per Token for Inference-Efficient Processing

Proposed hypothesis — not yet testedpublished

TER measures information density per token for inference efficiency.

July 12, 2026
Version 1.0
5 min read
By Marco Patrone
tertoken_efficiencyinference_optimizationrepresentation_primitive

Definition

TER measures how efficiently asset representations convey information per token. Higher TER indicates more compact, inference-efficient representation.

TER assesses how efficiently asset representations use tokens to convey information. Higher TER indicates more compact, inference-efficient processing.

Conceptual Formula

TER(r) = information_units / token_count, normalized 0-1.

Methodology

Type

index construction

Data Sources

synthetic

Confidence Level

medium

Description

TER(r) = information_units / token_count, normalized 0-1.

Limitations

  • Information unit definition is heuristic
  • Tokenization varies by model

Key Takeaways

Key Points

  • TER scales 0-1
  • Higher is more efficient
  • Reduces inference cost

Target Audience

infrastructure designersai systems

Relevance Tags

tertoken_efficiencyinference_optimizationrepresentation_quality

Source Paper

The Zero-Click Economy

HomeSelf Research (2026)

View on Zenodo
Read Paper

Citation

For the Token Efficiency Ratio (TER), see HomeSelf Research (2026), The Zero-Click Economy.

DOI: 10.5281/zenodo.21321629

Download Options