File size: 1,494 Bytes
459dab8 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 | ---
pretty_name: CP2Seq
tags:
- origami
- geometry
- spatial-reasoning
- synthetic
---
# CP2Seq
Synthetic origami crease patterns and reference folding sequences from
**[Leanmcp.com](https://leanmcp.com)**.
Authors: **Dheeraj Mohandas Pai and Lu Xian**.
Paper: *Origami as a Spatial and Geometric Reasoning Benchmark*.
Code, environment, and evaluation tools:
[Leanmcp-Community/CP2Seq](https://github.com/Leanmcp-Community/CP2Seq).
## Files
`data/` preserves the contents and relative paths of
`workspace/corpus/out/release/` in the source repository.
The release contains 600 samples across five batches:
| Batch | Samples |
| --- | ---: |
| all-layers | 400 |
| some-generated-d5 | 50 |
| some-generated-d6 | 50 |
| some-verified-d3 | 50 |
| some-verified-d4 | 50 |
Each sample directory contains `cp.fold`, `steps.fold`, `seq.json`, and
`meta.json`. The root `manifest.json` and `index.json` describe the release;
the check JSON files contain previously saved verification results.
The `verified` batch names refer to bounded search checks, not separate
training or test partitions. Reference sequences are not necessarily shortest.
This is a file-based geometry dataset, not a single tabular JSON dataset.
Download the original files with:
```sh
hf download Leanmcp/cp2seq --repo-type dataset --local-dir cp2seq
```
To use tools that expect the original repository layout, place the contents
of `cp2seq/data/` under `workspace/corpus/out/release/` in the code checkout.
|