OccTENS: 3D Occupancy World Model via Temporal Next-Scale Prediction

arXiv 25.9 2025 48.6 method

TLDR

OccTENS proposes a generative occupancy world model using temporal next-scale prediction for efficient, controllable long-term 3D scene generation.

Reasoning

The paper introduces a novel reformulation of occupancy world modeling as temporal next-scale prediction, addressing inefficiency, temporal degradation, and controllability. Strengths include a clear methodology (TensFormer, pose aggregation) and empirical outperformance over SOTA. Weaknesses: abstract lacks explicit mention of real-world datasets or benchmarks, though experiments imply standard evaluation.

Read-first score

Read-first score 48.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 36.

Recency 8%

86.7

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

Topical relevance 42%

51.4

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%

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

Reproducibility 25%

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

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 446.

Keyword Scores

world model

generative world model

world dynamics prediction

world simulator

video world model

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Reformulating occupancy world model as a temporal next-scale prediction (TENS) task, decomposing temporal sequence modeling into spatial scale-by-scale generation and temporal scene-by-scene prediction.
TensFormer architecture that effectively manages temporal causality and spatial relationships of occupancy sequences in a flexible and scalable way.
Holistic pose aggregation strategy for unified sequence modeling of occupancy and ego-motion, enhancing pose controllability.

Methodology

OccTENS reformulates the occupancy world model as a temporal next-scale prediction (TENS) task, decomposing temporal sequence modeling into spatial scale-by-scale generation and temporal scene-by-scene prediction using a TensFormer architecture. It also incorporates a holistic pose aggregation strategy for unified sequence modeling of occupancy and ego-motion to enhance controllability.

Key Results

OccTENS outperforms the state-of-the-art method with both higher occupancy quality and faster inference time.

Resources

Paper