Pixels to Proofs: Probabilistically-Safe Latent World Model Control via Parallel Conformal Robust MPC

arXiv 2026 56.7 method, application

TLDR

A framework for safe motion planning from pixels using robust MPC with conformal prediction in learned latent world models.

Reasoning

The paper introduces a novel combination of latent world models, conformal prediction, and robust MPC for safety, with strong empirical results on vision-based tasks. However, the abstract does not specify real-world validation, and the reliance on latent dynamics may limit interpretability.

Read-first score

Read-first score 56.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 49.

Recency 6%

100

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

Citation impact 18%

91.4

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.91440395

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=baseline

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: 456.

Keyword Scores

world model

world dynamics prediction

interactive world model

generative world model

world simulator

model-based reinforcement learning world model

video world model

Deep Analysis

Innovations

Training an action-conditioned joint-embedding world model with compact Markovian latent states for efficient gradient-based trajectory optimization.
Using conformal prediction to calibrate latent error bounds and inform a GPU-accelerated system level synthesis (SLS) robust MPC scheme.
Learning and conformalizing a latent constraint checker to impose probabilistic safety constraints during closed-loop execution.

Methodology

The framework trains an action-conditioned joint-embedding world model with compact Markovian latent states. It uses GPU-accelerated system level synthesis (SLS) robust MPC, informed by conformal prediction to obtain calibrated latent error bounds and robust latent-space constraint sets. Additionally, a latent constraint checker is learned and conformalized to enforce probabilistic safety constraints.

Key Results

On vision-based control tasks, the method improves both goal-reaching performance and safety over latent world-model and safe-planning baselines.

Resources

Paper