Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

GROKE: Vision-Free Navigation Instruction Evaluation via Graph Reasoning on OpenStreetMap

Farzad Shami, Subhrasankha Dey, Nico Van de Weghe, Henrikki Tenkanen

TL;DR

This work tackles the challenge of evaluating navigation instructions rather than navigation agents by proposing GROKE, a vision-free, training-free framework that reasons over OpenStreetMap graph data. It introduces a hierarchical, two-agent architecture (Sub-instruction Agent and Navigator Agent) and compares four spatial representations, finding that a structured JSON encoding provides the best balance of navigation accuracy and path fidelity. By treating agent execution success as a proxy for instruction navigability and correlating automated metrics with human judgments, GROKE offers a scalable, interpretable evaluation paradigm without visual data dependencies. The framework demonstrates meaningful improvements over baselines on Map2Seq data and outlines practical paths for future refinement, including model distillation and assistive-device integration.

Abstract

The evaluation of navigation instructions remains a persistent challenge in Vision-and-Language Navigation (VLN) research. Traditional reference-based metrics such as BLEU and ROUGE fail to capture the functional utility of spatial directives, specifically whether an instruction successfully guides a navigator to the intended destination. Although existing VLN agents could serve as evaluators, their reliance on high-fidelity visual simulators introduces licensing constraints and computational costs, and perception errors further confound linguistic quality assessment. This paper introduces GROKE(Graph-based Reasoning over OSM Knowledge for instruction Evaluation), a vision-free training-free hierarchical LLM-based framework for evaluating navigation instructions using OpenStreetMap data. Through systematic ablation studies, we demonstrate that structured JSON and textual formats for spatial information substantially outperform grid-based and visual graph representations. Our hierarchical architecture combines sub-instruction planning with topological graph navigation, reducing navigation error by 68.5% compared to heuristic and sampling baselines on the Map2Seq dataset. The agent's execution success, trajectory fidelity, and decision patterns serve as proxy metrics for functional navigability given OSM-visible landmarks and topology, establishing a scalable and interpretable evaluation paradigm without visual dependencies. Code and data are available at https://anonymous.4open.science/r/groke.

GROKE: Vision-Free Navigation Instruction Evaluation via Graph Reasoning on OpenStreetMap

TL;DR

This work tackles the challenge of evaluating navigation instructions rather than navigation agents by proposing GROKE, a vision-free, training-free framework that reasons over OpenStreetMap graph data. It introduces a hierarchical, two-agent architecture (Sub-instruction Agent and Navigator Agent) and compares four spatial representations, finding that a structured JSON encoding provides the best balance of navigation accuracy and path fidelity. By treating agent execution success as a proxy for instruction navigability and correlating automated metrics with human judgments, GROKE offers a scalable, interpretable evaluation paradigm without visual data dependencies. The framework demonstrates meaningful improvements over baselines on Map2Seq data and outlines practical paths for future refinement, including model distillation and assistive-device integration.

Abstract

The evaluation of navigation instructions remains a persistent challenge in Vision-and-Language Navigation (VLN) research. Traditional reference-based metrics such as BLEU and ROUGE fail to capture the functional utility of spatial directives, specifically whether an instruction successfully guides a navigator to the intended destination. Although existing VLN agents could serve as evaluators, their reliance on high-fidelity visual simulators introduces licensing constraints and computational costs, and perception errors further confound linguistic quality assessment. This paper introduces GROKE(Graph-based Reasoning over OSM Knowledge for instruction Evaluation), a vision-free training-free hierarchical LLM-based framework for evaluating navigation instructions using OpenStreetMap data. Through systematic ablation studies, we demonstrate that structured JSON and textual formats for spatial information substantially outperform grid-based and visual graph representations. Our hierarchical architecture combines sub-instruction planning with topological graph navigation, reducing navigation error by 68.5% compared to heuristic and sampling baselines on the Map2Seq dataset. The agent's execution success, trajectory fidelity, and decision patterns serve as proxy metrics for functional navigability given OSM-visible landmarks and topology, establishing a scalable and interpretable evaluation paradigm without visual dependencies. Code and data are available at https://anonymous.4open.science/r/groke.
Paper Structure (42 sections, 10 equations, 7 figures, 10 tables, 1 algorithm)

This paper contains 42 sections, 10 equations, 7 figures, 10 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Evaluation Deficits in Embodied Systems.
  3. Visual Reliance in Pragmatic Evaluation.
  4. Research Objectives.
  5. Contribution.
  6. Background & Related Work
  7. Transition from Vision-Dependent to Graph-Based Navigation.
  8. Spatial Representation for Language Model Reasoning
  9. Agent-Based Evaluation and Metric Evolution
  10. Methodology
  11. Overview
  12. Problem Formulation.
  13. System Architecture.
  14. Sub-instruction Agent
  15. Sub-goal Decomposition.
  16. ...and 27 more sections

Figures (7)

  • Figure 1: Comparison of traditional textual metrics and proposed pragmatic evaluation metrics.
  • Figure 2: (Left) consists of three modules: Sub-Goal & POI Extraction, Visible Area Construction, and Navigator Agent. These modules generate sub-goal instructions, construct spatial information representations for the surrounding area, and traverse the graph, respectively. (Right) Visible Area Construction in detail: The map highlights the distinction between different kinds of nodes for immediate local context construction. It demonstrates how the system filters out distant data using visibility thresholds to construct the immediate navigable context.
  • Figure 3: Normalized complexity score distribution.
  • Figure 4: Matrix / grid representation for the second iteration of the instruction: "Make two left turns to head the opposite way on the other side of the block." The first left turn is already executed, as indicated by the past trajectory denoted as '1' in the representation.
  • Figure 5: Overall performance comparison on Navigation Execution for different types of spatial information representation.
  • ...and 2 more figures