⌁ Awesome World Model Hub Papers · Datasets · Projects
中文 GitHub
← Back to papers

Latent World Models for Automated Driving: A Unified Taxonomy, Evaluation Framework, and Open Challenges

arXiv 26.3 2026 71.2 survey, application

TLDR

A unifying taxonomy and evaluation framework for latent world models in automated driving, covering design space, internal mechanics, and open challenges.

Reasoning

The paper provides a comprehensive synthesis of latent world models for driving, with a clear taxonomy and evaluation prescriptions. However, it lacks real-world experiments or empirical validation, being a survey/framework paper.

Read-first score

Read-first score 71.2, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 51.

Methodology quality 25%
100

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

Recency 8%
100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Topical relevance 42%
72.9

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%
30

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

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 61.

Keyword Scores

world model
10
generative world model
9
world dynamics prediction
8
world simulator
7
interactive world model
6
model-based reinforcement learning world model
6
video world model
5

Deep Analysis

Innovations

  • Unified latent-space framework for world models in automated driving
  • Taxonomy organizing design space by latent representations (latent worlds, latent actions, latent generators; continuous states, discrete tokens, hybrids) and structural priors (geometry, topology, semantics)
  • Identification of five cross-cutting internal mechanics: structural isomorphism, long-horizon temporal stability, semantic and reasoning alignment, value-aligned objectives and post-training, adaptive computation and deliberation
  • Proposal of a closed-loop metric suite and a resource-aware deliberation cost to reduce open-loop/closed-loop mismatch
  • Actionable research directions toward decision-ready, verifiable, and resource-efficient automated driving

Methodology

The paper synthesizes recent progress in world models for automated driving by proposing a unifying latent-space framework. It organizes the design space according to the target and form of latent representations and structural priors, and articulates five cross-cutting internal mechanics. It also proposes evaluation prescriptions including a closed-loop metric suite and a resource-aware deliberation cost, though no experimental implementation is described.

Key Results

The paper presents a conceptual taxonomy and evaluation framework for latent world models in automated driving, but does not provide experimental results or empirical validation of the proposed ideas.

Limitations

  • The proposed framework is theoretical and lacks empirical validation
  • The taxonomy may not cover all existing approaches or future developments
  • The evaluation prescriptions are proposed but not implemented or tested
  • The paper does not present new experimental results or comparisons with baselines

Resources

Paper

Tags