RoleWorld: Knowledge Transfer in World Modeling with Functional Role Abstractions

arXiv 2026 35.4 method

TLDR

RoleWorld introduces symbolic world modeling with reusable functional role abstractions and executable transition rules, enabling LLM agents to transfer knowledge across environments, validated on MARS benchmark.

Reasoning

The paper presents a novel symbolic approach for world model transfer using role prototypes and parameterized rules, with strong empirical validation on MARS. However, the abstract lacks details on baseline comparisons and potential limitations of fixed transition rules in highly novel environments.

Read-first score

Read-first score 35.4, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 34.

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=benchmark,evaluation,experiment

Topical relevance 25%

48.6

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,github

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: 427.

Keyword Scores

world model

world dynamics prediction

interactive world model

world simulator

model-based reinforcement learning world model

generative world model

video world model

Deep Analysis

Innovations

Introduces functional role abstractions that combine reusable prototypes with environment-specific parameter values to represent object behavior, instead of organizing symbolic world knowledge around environment-specific objects.
A teacher LLM extracts role prototypes, executable transition rules defined over role instances, and a skill that guides the actor LLM to use role information in decision making.
In a target world, the actor uses the skill with current role assignments and parameter estimates, actively testing role-instance hypotheses through interaction when uncertainty affects decisions.
Role assignments and parameter estimates are refined online from action feedback without LLM parameter updates, while transition rules remain fixed.

Methodology

RoleWorld is a symbolic world modeling approach where a teacher LLM extracts role prototypes, executable transition rules over role instances, and a skill from interaction trajectories. In target environments, the actor LLM uses this skill with current role assignments and parameter estimates, actively testing uncertain role hypotheses during decision making and updating role information online from feedback. Evaluation is performed on the MARS benchmark without updating LLM parameters and with transition rules held fixed.

Key Results

Experiments on the MARS benchmark show that RoleWorld enables robust generalization across environments with significantly altered object behavior.

Resources

Paper