GEM: Generating LiDAR World Model via Deformable Mamba

arXiv 2026 61.3 method

TLDR

GEM uses deformable Mamba to generate LiDAR world models, addressing point cloud disorder and dynamic-static separation for autonomous driving.

Reasoning

The paper introduces a novel architecture (deformable Mamba) for LiDAR world modeling, with strong empirical results on benchmarks. However, the abstract lacks details on limitations and comparison to baselines, and the optional planner/controller is not fully evaluated.

Read-first score

Read-first score 61.3, 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

Reproducibility 18%

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

Citation impact 18%

71.1

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.71079333

Methodology quality 18%

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

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

Citation velocity 12%

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

Field roles

FrontierBridgeMethodology anchorReproducibility anchor

Rank sensitivity

Stability: volatile; rank range: 461.

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

Deformable Mamba architecture for LiDAR world model
Custom LiDAR scene tokenizer to tokenize LiDAR sweeps into compact representations
Unsupervised dynamic-static separator for disentangling tokenized features
Tri-path deformable Mamba with selective scanning and adaptive gating fusion
Integration of planner and BEV layout controller for autonomous rollout and 'what-if' scenarios

Methodology

The method tokenizes LiDAR sweeps into compact representations using a custom LiDAR scene tokenizer, then performs unsupervised disentanglement of tokenized features via a dynamic-static separator. A tri-path deformable Mamba is introduced to perform selective scanning and adaptive gating fusion over the disentangled features, enhancing spatial-temporal understanding of world evolution.

Key Results

GEM achieves state-of-the-art performances across diverse benchmarks and evaluation settings, demonstrating its superiority and effectiveness in LiDAR world modeling.

Resources

Paper Code