3D-Belief: Embodied Belief Inference via Generative 3D World Modeling

arXiv 2026 49.2 method

TLDR

Proposes 3D-Belief, a generative 3D world model for embodied agents to infer and update beliefs about unobserved 3D space from partial observations.

Reasoning

The paper introduces a novel perspective on world modeling as embodied belief inference in 3D, with explicit uncertainty representation and online updating. Strengths include a clear conceptual shift from visual realism to structured uncertainty and real-world validation; weaknesses are the narrow focus on 3D scenes and lack of comparison to alternative world model paradigms.

Read-first score

Read-first score 49.2, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 33.

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=0.76681221

Methodology quality 18%

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

Topical relevance 29%

47.1

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=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: 288.

Keyword Scores

world model

generative world model

interactive world model

world simulator

world dynamics prediction

model-based reinforcement learning world model

video world model

Deep Analysis

Innovations

World modeling as embodied belief inference in 3D space, maintaining and updating an agent's belief about unobserved 3D world
Spatially consistent scene memory for partial observations
Multi-hypothesis belief sampling to represent uncertainty in 3D
Sequential belief updating over time
Semantically informed prediction of unseen regions
Generative 3D world model that infers explicit, actionable 3D beliefs and updates them online

Methodology

3D-Belief is a generative 3D world model that infers explicit 3D beliefs from partial observations and updates them online. It incorporates spatially consistent scene memory, multi-hypothesis belief sampling, sequential belief updating, and semantically informed prediction of unseen regions. The model is evaluated on 2D visual quality, 3D imagination using the proposed 3D-CORE benchmark, and object navigation tasks in both simulation and the real world, comparing against state-of-the-art methods.

Key Results

3D-Belief improves 2D and 3D imagination quality and downstream embodied task performance (object navigation) compared to state-of-the-art methods.

Resources

Paper