ActionRoPE: action as coordinates for spatially consistent control in planar-camera game world models

arXiv 2026 36.7 method

TLDR

Introduces ActionRoPE, a parameter-free action interface mapping camera displacement to world-coordinate rotary positions, plus Coordinate-Conditioned Noise Scheduling, for spatially consistent control in planar-camera game world models.

Reasoning

The paper proposes a novel action interface and noise scheduling approach to improve spatial consistency in game world models, with strong conceptual grounding in world-coordinate geometry and an empirical comparison showing a 66% reduction in revisit drift. Weaknesses include a narrow focus on planar-camera games and limited abstract-level detail on broader generalization or benchmark scope.

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=baseline,metric,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

generative world model

interactive world model

world simulator

video world model

world dynamics prediction

model-based reinforcement learning world model

Deep Analysis

Innovations

ActionRoPE: a parameter-free action interface that converts action-driven planar camera displacement into world-coordinate rotary positions, so tokens corresponding to the same world location retain the same positional address across time.
Coordinate-Conditioned Noise Scheduling (CNS): token-wise denoising schedules based on the known-new spatial asymmetry, allowing observed regions to stabilize earlier while newly revealed regions are synthesized with reference to them.
Identification of a spatial asymmetry in planar-camera generation: regions covered by the initial observation are geometrically known, while newly revealed regions remain to be synthesized.

Methodology

ActionRoPE represents actions as planar camera displacements and encodes them directly as world-coordinate rotary positions rather than using feature conditioning. Coordinate-Conditioned Noise Scheduling assigns token-wise denoising schedules according to whether tokens correspond to known observed regions or newly revealed regions. The method is evaluated against four representative action-conditioning mechanisms using video backbones from 1.3B to 33B parameters, including out-of-distribution scenes, with metrics of commanded-motion faithfulness and revisit drift.

Key Results

ActionRoPE follows commanded motion more faithfully than the compared action-conditioning mechanisms and reduces revisit drift by 66% over the strongest baseline. The same interface transfers across video backbones from 1.3B to 33B parameters and to out-of-distribution scenes.

Resources

Paper