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

TARDIS STRIDE: A Spatio-Temporal Road Image Dataset and World Model for Autonomy

arXiv 2025 71.8 method, benchmark, application

TLDR

Introduces STRIDE, a spatio-temporal road image dataset, and TARDIS, a generative world model for autonomous agents with spatial-temporal dynamics.

Reasoning

The paper presents a strong dataset and world model with broad agentic evaluations, though only title/abstract were visible. Weaknesses include lack of detailed experimental methodology and limited evidence for real-world deployment claims.

Read-first score

Read-first score 71.8, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 58.

Recency 8%
86.7

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

Topical relevance 42%
82.9

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

Methodology quality 25%
70

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

Reproducibility 25%
50

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

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 79.

Keyword Scores

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

Deep Analysis

Innovations

  • STRIDE dataset: a spatio-temporal road image dataset that permutes 360-degree panoramic imagery into interconnected observation, state, and action nodes.
  • TARDIS model: a transformer-based generative world model that integrates spatial and temporal dynamics through a unified autoregressive framework.
  • Unified modeling of egocentric views, positional coordinates, and movement commands across both space and time for autonomous agent tasks.

Methodology

STRIDE is constructed by permuting 360-degree panoramic imagery into interconnected observation, state, and action nodes, capturing relationships among egocentric views, positional coordinates, and movement commands across space and time. TARDIS is a transformer-based generative world model trained on STRIDE using a unified autoregressive framework that jointly models spatial and temporal dynamics. The approach is benchmarked on controllable photorealistic image synthesis, instruction following, autonomous self-control, and georeferencing tasks.

Key Results

TARDIS demonstrates robust performance across controllable photorealistic image synthesis, instruction following, and autonomous self-control, and achieves state-of-the-art georeferencing. These results indicate progress toward generalist agents with enhanced embodied spatial-temporal reasoning.

Resources

Paper

Tags