UniDrive-WM: Unified Understanding, Planning and Generation World Model For Autonomous Driving

arXiv 26.1 2026 42.6 method, application

TLDR

UniDrive-WM is a unified VLM-based world model for autonomous driving that jointly performs scene understanding, trajectory planning, and future image generation.

Reasoning

The paper's strength lies in tightly integrating perception, planning, and generation within a single VLM architecture, achieving significant improvements on the Bench2Drive benchmark. However, it is limited to a single benchmark and lacks real-world deployment or diverse scenario validation.

Read-first score

Read-first score 42.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 42.

Recency 6%

100

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

Methodology quality 18%

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

Topical relevance 29%

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 18%

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

Citation impact 18%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 329.

Keyword Scores

world model

generative world model

world dynamics prediction

world simulator

video world model

model-based reinforcement learning world model

interactive world model

Deep Analysis

Innovations

Unified VLM-based world model that jointly performs driving-scene understanding, trajectory planning, and trajectory-conditioned future image generation within a single architecture
Trajectory planner predicts future trajectory which conditions a VLM-based image generator to produce plausible future frames, providing additional supervisory signals to enhance scene understanding and iteratively refine trajectory generation
Comparison of discrete and continuous output representations for future image prediction and their influence on downstream driving performance

Methodology

UniDrive-WM is a unified VLM-based world model that jointly performs scene understanding, trajectory planning, and trajectory-conditioned future image generation. The trajectory planner predicts a future trajectory, which conditions a VLM-based image generator to produce future frames; these predictions provide supervisory signals to enhance understanding and iteratively refine trajectory generation. The model is evaluated on the Bench2Drive benchmark.

Key Results

UniDrive-WM improves planning performance by 7.3% in L2 trajectory error and 10.4% in collision rate over the previous best method on the Bench2Drive benchmark, while producing high-fidelity future images.

Resources

Paper Project