Dataset Viewer
Auto-converted to Parquet Duplicate
Search is not available for this dataset
The dataset viewer is not available for this split.
Server error while post-processing the rows. This occured on row 11. Please report the issue.
Error code:   RowsPostProcessingError

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

World Cellular Automata Technical Report v0.1

This release contains the public technical report package for World Cellular Automata (WCA): a recursive local-world architecture for world-state modeling.

Related Public Repositories

What WCA Is

WCA represents a world state as node vectors, expands that state into a family of node-centered local worlds, evolves those local worlds through learned pair interactions, and recomposes evolved centers into the next global state.

The core semantic contract is:

H_t [B,N,D]
  -> L_t [B,N,N,D]
  -> dense local-world pair evolution with semantic [B,N,N,N,D]
  -> center recomposition
  -> H_{t+1} [B,N,D]

Release Contents

  • wca_technical_report_en.pdf: English technical report.
  • wca_technical_report_zh.pdf: Chinese technical report.
  • latex/: LaTeX sources.
  • figures/: Figures used by the report.
  • tables/: Main result table.
  • model_card.md: WCA architecture/model card.
  • dataset_and_artifact_card.md: Dataset, split, artifact, and evaluation-card summary.
  • zenodo_metadata.json: Draft Zenodo metadata.
  • CITATION.cff: Citation metadata template.
  • release_manifest.json: File hashes for this release package.

Main Claim

WCA is a credible early architecture primitive for recursive world-state modeling. Current evidence is strongest in token-level PDEBench reaction-diffusion, where recursion depth, learnable interfaces, attribution controls, and h8x2 rollout diagnostics support WCA-core contribution beyond interface-only and no-recursion controls.

Evidence Status

  • Formal current row: V25d h8x2 token-level rollout.
  • Source-audit attribution: V25e-r3 learnable-interface controls.
  • Establish-stage evidence: V25 recursion-depth ladder.
  • Secondary evidence: maze potential fields and WeatherBench-style real-field transfer.
  • Diagnostic boundary: N=256 PatchMean underperformance.

Recommended Citation

Use the metadata in CITATION.cff. Replace placeholder author fields before formal archive submission if individual authorship should be listed.

Archive And DOI

Use the Zenodo DOI for archival citation: https://doi.org/10.5281/zenodo.21085154.

Record page: https://zenodo.org/records/21085154.

For reproducibility, cite both the technical report DOI and the source-code commit or release tag used to run experiments.

Downloads last month
158