World-Task Factorization for Robot Learning

arXiv 2026 36.7 method, theory

TLDR

Proposes world-task factorization for robot learning, separating environment properties from task logic using Bayesian evidence and differentiable graphs.

Reasoning

Strengths include a novel factorization framework and formalization via Bayesian model evidence. Weaknesses are the lack of empirical validation or real-world experiments in the abstract, making the practical impact unclear.

Read-first score

Read-first score 36.7, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 4.

Recency 6%

100

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

Citation impact 18%

85.3

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.85277201

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

Topical relevance 29%

5.7

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

Citation velocity 12%

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

Field roles

FoundationFrontier

Rank sensitivity

Stability: volatile; rank range: 273.

Keyword Scores

world model

world dynamics prediction

model-based reinforcement learning world model

world simulator

generative world model

interactive world model

video world model

Deep Analysis

Innovations

Proposing world-task factorization as the most fundamental factorization in robotics, separating world factors (properties of embodied system and environment, independent of intent) from task factors (task logic over what the world admits).
Formalizing the asymmetry through Bayesian model evidence, aligning with the data-generating process, maintaining high likelihood via an analytical world model, and reducing the Occam's razor penalty on task parameters.
Instantiating the factorization with AICON (a differentiable graph of recursive estimators and interconnections) that is compositional, operates without task-specific data, and propagates cost gradients to actuators, paired with a compact learned policy that modulates gradient paths, using gradients as the interface between world and task factors.

Methodology

The framework pairs AICON, a differentiable graph of recursive estimators and interconnections that is compositional and operates without task-specific data, with a compact learned policy that modulates gradient paths. Gradients serve as the interface: they carry world structure through the graph and task structure through costs, enabling low-dimensional learning while preserving structural generalization. The approach is tested across three problems encompassing heterogeneous robots, environments, task logic, and sensorimotor modalities, compared against end-to-end baselines and analytical heuristics.

Key Results

The framework outperforms end-to-end baselines and analytical heuristics in all settings, generalizes zero-shot to out-of-distribution configurations, and transfers to real hardware without retraining.

Resources

Paper