WorldStream: Streaming Interactive 3D Worlds with Hierarchical Planning and Persistent State

arXiv 2026 35.3 method

TLDR

WorldStream is an open-ended 3D world model that streams infinite terrain using hierarchical planning and persistent state, enabling language interaction and restoration of edited locations.

Reasoning

The paper proposes a novel system for streaming interactive 3D worlds with hierarchical planning and persistent state, which is conceptually strong. However, the abstract provides no empirical evaluation or benchmark results, so the practical effectiveness and limitations are not evidenced.

Read-first score

Read-first score 35.3, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 33.

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,experiment,result

Topical relevance 25%

47.1

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: 461.

Keyword Scores

world model

interactive world model

generative world model

world simulator

world dynamics prediction

video world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Open-ended 3D world model that simultaneously extends playable terrain beyond any pre-specified map, acts on open-ended language in the running world, and restores visited places with their edits.
Hierarchical planner over a typed decision model that emits structured decision state, with Macro fixing settlement, ecology, transport, and hydrological structure across tiles and Micro specializing tile semantics under terrain and neighbor commitments.
Deterministic local code that compiles accepted plans into terrain, roads, settlements, vegetation, and entities.
Streaming architecture that separates planning, compilation, and geometry retention: plans run ahead along the predicted corridor, compilation occurs only where movement will soon need it, and behind-user geometry is released.
Persistent world record retaining plans, seeds, object identities and mutable state, boundary commitments, and per-attribute edits, revised as the user advances and edits; returning tiles recompile from the record.
Interaction grounded in the persistent record, enabling the model to act on existing entities, update stateful NPCs as they act, edit the present region, or attach intent to unplanned regions.

Methodology

WorldStream uses a hierarchical planner over a typed decision model to emit structured Macro and Micro decision state. Accepted plans are compiled by deterministic local code into terrain, roads, settlements, vegetation, and entities. A streaming design runs plans ahead, compiles near upcoming movement, releases behind-user geometry, and a persistent world record stores plans, seeds, identities, mutable state, boundary commitments, and edits for recompilation and interaction.

Key Results

The abstract reports no quantitative experimental results or baseline comparisons; it states that the combined system yields a world expanding under bounded budgets, connected across independently planned tiles, and interactive.

Limitations

No limitations are explicitly stated in the title or abstract.

Resources

Paper