GEM: Gaussian Evolution Model for Occupancy Forecasting and Motion Planning

arXiv 2026 52.9 method, application

TLDR

GEM uses explicit 4D Gaussian primitives for non-autoregressive occupancy forecasting and motion planning in autonomous driving.

Reasoning

The paper introduces a novel continuous-time representation that avoids autoregressive error accumulation, achieving state-of-the-art results. However, the abstract is cut off, and the method is domain-specific to autonomous driving, limiting generalizability.

Read-first score

Read-first score 52.9, 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

Citation impact 18%

76.7

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.76663262

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

Methodology quality 18%

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

Reproducibility 18%

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

Citation velocity 12%

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

Field roles

FoundationFrontierBridge

Rank sensitivity

Stability: volatile; rank range: 282.

Keyword Scores

world model

world dynamics prediction

generative world model

video world model

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Non-autoregressive occupancy world modeling using explicit continuous 4D Gaussian primitives with learned dynamics
Direct querying of the Gaussian world representation at arbitrary timestamps for flexible temporal forecasting
Decoupling spatial geometry, temporal support, and primitive motion for interpretable scene evolution
Unified representation supporting both future semantic occupancy forecasting and motion planning

Methodology

GEM represents driving scenes as explicit continuous 4D Gaussian primitives with learned dynamics. Instead of autoregressive step-by-step forecasting, it directly queries the Gaussian world at arbitrary timestamps and splats the corresponding conditional 3D Gaussians into semantic occupancy volumes, enabling efficient non-autoregressive forecasting over the full horizon.

Key Results

GEM achieves state-of-the-art future semantic occupancy forecasting and strong motion planning performance, while providing flexible temporal querying.

Resources

Paper