fhyfhy/diffusion-vs-ar-hard-sudoku

Diffusion vs AR Hard Sudoku

This repository packages 8,148,696 Sudoku examples in the CSV format expected by HKUNLP/diffusion-vs-ar, plus its original 100k/1k easy baseline.

Every processed file has these columns:

column	meaning
quizzes	81 row-major digits; 0 is an empty cell
solutions	complete 81-digit solution
source	original collection
dataset	normalized dataset family
official_rating	rating supplied by the source
rating_type	semantics of that rating
difficulty_bucket	categorical bucket used for mixtures
clues	number of givens
split	source-preserving train/test designation

Dataset families

• Sudoku Extreme (4,254,780): tdoku_backtracks; official train/test split is preserved. Buckets: r0, r1_4, r5_19, r20_49, r50_99, r100_plus.
• Radcliffe Kaggle 3M (3,000,000): mean_search_depth_10_runs. Buckets: r0, r1_2, r2_4, r4_plus.
• Sudoku Exchange (893,916): sukaku_explainer. Buckets: easy, medium, hard, diabolical. Solutions were generated with a bitmask MRV solver; the source states that all puzzles have a unique solution.

Raw ratings from different families are not numerically comparable.

Licensing and provenance

• Sudoku Exchange dedicates its puzzle bank to the public domain.
• Radcliffe's 3M dataset is published as CC0.
• Sudoku Extreme combines several community benchmark sources and does not declare one uniform license in its data card. Review its listed upstream sources before redistribution outside research use.

See processed/manifest.json for exact per-file counts. Conversion, validation, training, and multi-difficulty evaluation code is available in the hard-sudoku-datasets GitHub branch.