Multiple datasets for ARC (Abstraction and Reasoning Corpus)

Each dataset have its own license, see the readme/license for details.

The datasets can be inspected here: ARC-Interactive

---

Updates in This Fork

This fork of the arc-dataset-collection repository includes the following modifications:

• Latest Versions Only: Replaced all older dataset versions with their latest available versions.
• Unified Naming: Removed version suffixes (e.g., _v2, _v3) from file and directory names for clarity and simplicity. All files now reflect the most up-to-date version by default.

997M -> 952M: delete blank space in json.

952M -> 830M: packing arc-heavy to jsonl and compress to zip.