PeytonT/paper_universe_interactive

---
pretty_name: Paper Universe Interactive Graph
viewer: true
tags:
- datasets
- graph
- scientific-papers
- arxiv
- webgl
- wasm
- visualization
size_categories:
- 10M<n<100M
configs:
- config_name: papers_50k
data_files:
- split: train
path: parquet/interactive/papers_50000.parquet
- config_name: papers_200k
data_files:
- split: train
path: parquet/interactive/papers_200000.parquet
- config_name: papers_all
data_files:
- split: train
path: parquet/interactive/papers_all.parquet
- config_name: categories
data_files:
- split: train
path: parquet/interactive/categories.parquet
- config_name: years
data_files:
- split: train
path: parquet/interactive/years.parquet
---

# Paper Universe Interactive Graph

Small static-viewer payload for the Research Library paper universe.

This dataset is intentionally separate from `PeytonT/paper_graph`. It contains browser-friendly interactive assets needed by the static WebGL/WASM app. Parquet is the preferred payload format; JSON is retained as a compatibility fallback:

- `parquet/interactive/papers_50000.parquet`
- `parquet/interactive/papers_200000.parquet`
- `parquet/interactive/papers_all.parquet`
- `parquet/interactive/categories.parquet`
- `parquet/interactive/years.parquet`
- fallback JSON under `interactive/`
- `interactive/manifest.json`
- generated HTML viewers
- PNG overview renders
- lightweight build/progress manifests

It does **not** include full parquet graph splits, paper embeddings, full-text embeddings, KNN edges, topic nodes, or `papers_all.json`.

Selected paper views should compute nearest-neighbor context from the loaded interactive level in the browser. The full paper KNN parquet remains in `PeytonT/paper_graph` and is intentionally excluded from this lightweight static payload.

## Size

- parquet payload bytes: about `68.2 MiB`
- approximate total with JSON fallback and renders: `138 MiB`
- largest preferred payload: `parquet/interactive/papers_all.parquet`

## Intended Use

Static app URL pattern:

```text
https://huggingface.co/datasets/PeytonT/paper_universe_interactive/resolve/main/interactive/manifest.json
```

The static `research_library` app prefers the Parquet paths in `interactive/manifest.json` through DuckDB-WASM, caches them client-side, and uses WASM to normalize coordinate buffers before WebGL rendering. If Parquet loading is unavailable, it falls back to the JSON files.

## Relationship To Full Dataset

Use `PeytonT/paper_graph` for the full research graph and embeddings. Use this dataset for fast static browser visualization.