Awesome World Model Hub Papers · Datasets · Projects
← Back to papers

SURGE: Approximation and Training Free Particle Filter for Diffusion Surrogate

arXiv 2026 49.5 method

TLDR

A particle filter method using diffusion models as world models for data assimilation, with Sequential Monte Carlo correction.

Reasoning

The paper presents a novel integration of diffusion models with particle filtering for unbiased posterior sampling in data assimilation. Its strength lies in the rigorous theoretical framework, but it lacks real-world experimental validation, limiting its immediate practical impact.

Read-first score

Read-first score 49.5, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 36.

Recency 6%
100

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

Citation impact 18%
71.3

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.71301326

Methodology quality 18%
60

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

Topical relevance 29%
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

Reproducibility 18%
30

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

Citation velocity 12%
0

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

Field roles

FrontierBridge

Rank sensitivity

Stability: volatile; rank range: 244.

Keyword Scores

world model
10
generative world model
9
world dynamics prediction
9
world simulator
8
interactive world model
0
video world model
0
model-based reinforcement learning world model
0

Deep Analysis

Innovations

  • Combining diffusion models with particle filtering for data assimilation without approximation or training
  • Sequential Monte Carlo over the diffusion trajectory viewed as a path measure to reweight and resample particles
  • Unbiased particle filtering method that rigorously fuses observational data with diffusion model simulations

Methodology

The method employs a diffusion model as a world model to simulate system dynamics. It represents the posterior distribution using a set of particles, guides the diffusion generation process with observation likelihood, and then applies Sequential Monte Carlo over the diffusion trajectory to reweight and resample particles, ensuring convergence to the true posterior.

Key Results

The abstract claims the method leads to an unbiased particle filtering approach that rigorously fuses observational data with diffusion model simulations, but no experimental results are presented.

Tags

data assimilationdiffusion modelsparticle filterstate estimationscore-based modelssequential inferenceMLLG