ATLAS: Aligned Transport of Latent Structure for Reliable World Model Planning

arXiv 2026 36.7 method

TLDR

ATLAS preserves relational geometry in latent world models for more reliable planning, improving goal-reaching success on PushT, TwoRoom, and OGBench-Cube.

Reasoning

The paper clearly motivates and validates a novel training objective for latent world models, with strong empirical evaluation across multiple benchmarks. Weaknesses are that the abstract lacks details on baselines, computational cost, and limitations, and the contribution is somewhat narrow relative to broader world-model framing.

Read-first score

Read-first score 36.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 41.

Recency 5%

100

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

Methodology quality 15%

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

Topical relevance 25%

58.6

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 15%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=code

Graph prestige 15%

5.7

PageRank-style prestige over the local citation graph built from available OpenAlex references. local_pagerank=0.0016

Citation impact 15%

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. cited_by_count=0

Citation velocity 10%

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

Field roles

FrontierMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 362.

Keyword Scores

world model

model-based reinforcement learning world model

world dynamics prediction

world simulator

interactive world model

generative world model

video world model

Deep Analysis

Innovations

Identifies that regularizing the latent marginal alone does not determine state-to-state relationships, allowing planning-relevant novelty structure to be weakened in the planner's final latent.
Introduces ATLAS, a training objective that preserves relational geometry while calibrating the global latent distribution.
Transfers normalized pairwise structure from an informative encoder representation to the planning latent.
Uses Wasserstein embedding matching (WEMReg) to calibrate the planning latent marginal via one-dimensional Wasserstein-2 transport.
Provides analysis showing relational preservation and marginal calibration are non-redundant and linking finite-candidate planning stability to relational distortion, latent-scale mismatch, and prediction error.

Methodology

ATLAS is instantiated in LeWM and combines preservation of normalized pairwise relational structure from an encoder representation with Wasserstein embedding matching for marginal calibration. Evaluation uses goal-reaching success on PushT, TwoRoom, and OGBench-Cube under lower- and higher-novelty subsets, plus representation and rollout diagnostics for novelty structure, marginal calibration, and multi-step prediction error.

Key Results

ATLAS improves mean goal-reaching success across PushT, TwoRoom, and OGBench-Cube on both lower- and higher-novelty evaluation subsets, with the largest gain on higher-novelty TwoRoom episodes; diagnostics show stronger novelty-related structure, improved marginal calibration, and lower multi-step prediction error.

Resources

Paper