NavMorph: A Self-Evolving World Model for Vision-and-Language Navigation in Continuous Environments

ICCV 25 2025 57.3 method, application

TLDR

NavMorph is a self-evolving world model for VLN-CE that uses latent representations and contextual memory to improve navigation adaptability.

Reasoning

The paper introduces a novel framework that addresses generalization and adaptation in VLN-CE, with clear methodology and benchmark results. However, the abstract lacks explicit discussion of limitations or comparisons to baselines, and the experiments are in simulated environments, not real-world.

Read-first score

Read-first score 57.3, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 25.

Recency 8%

86.7

Uses a gentle age decay so recent papers surface without erasing older foundations. 2025

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=True; dataset=False; markers=code,github

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,experiment

Topical relevance 42%

35.7

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Field roles

FrontierReproducibility anchor

Rank sensitivity

Stability: volatile; rank range: 515.

Keyword Scores

world model

world dynamics prediction

model-based reinforcement learning world model

interactive world model

world simulator

generative world model

video world model

Deep Analysis

Innovations

Self-evolving world model framework for VLN-CE that enhances environmental understanding and decision-making
Compact latent representations to model environmental dynamics, enabling foresight for adaptive planning and policy refinement
Contextual Evolution Memory that leverages scene-contextual information for effective navigation and online adaptability

Methodology

NavMorph is a self-evolving world model that uses compact latent representations to model environmental dynamics, integrating a Contextual Evolution Memory to leverage scene-contextual information for adaptive planning and policy refinement. The model is trained and evaluated on VLN-CE benchmarks, with code publicly available.

Key Results

The method achieves notable performance improvements on popular VLN-CE benchmarks, demonstrating enhanced generalization and adaptability.

Resources

Paper Code