NavForesee: A Unified Vision-Language World Model for Hierarchical Planning and Dual-Horizon Navigation Prediction

arXiv 25.12 2025 46.4 method, application

TLDR

NavForesee unifies language planning and generative world model prediction for embodied navigation, achieving competitive results on R2R-CE and RxR-CE.

Reasoning

Strengths: novel integration of planning and prediction in a single VLM with internal feedback loop. Weaknesses: limited to navigation tasks, no explicit comparison to model-based RL world models, abstract lacks discussion of limitations.

Read-first score

Read-first score 46.4, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 28.

Recency 8%

86.7

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

Methodology quality 25%

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

Topical relevance 42%

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

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 376.

Keyword Scores

world model

generative world model

world dynamics prediction

world simulator

interactive world model

video world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Unified Vision-Language Model (VLM) that concurrently performs high-level language planning and predictive world model imagination within a single framework
Dual-horizon prediction: short-term environmental dynamics and long-term navigation milestones
Internal feedback loop of perception-planning/prediction-action where structured plan guides targeted prediction and imagined future informs navigation actions

Methodology

NavForesee is a VLM conditioned on the full instruction and historical observations. It is trained to decompose the task, track progress, and formulate subsequent sub-goals while simultaneously functioning as a generative world model that predicts short-term environmental dynamics and long-term navigation milestones. The model creates a feedback loop where the plan guides prediction and imagined future informs actions.

Key Results

NavForesee achieves highly competitive performance on the R2R-CE and RxR-CE benchmarks in complex long-horizon navigation scenarios.

Resources

Paper