Tool
Open catalog excerpts, not the files
Only excerpts already stored on the catalog row are shown. The count is the length of the samples array. GearDel does not host archives. Quality checks run on those excerpts only.
Excerpts are already on the public catalog row. Pasting your own file still stays in this tab and is not uploaded.
What you are looking at
The list is catalog rows whose samples array is not empty. The number next to a name is that array’s length, not a file size or download count.
Selecting a row shows those excerpts and runs the same quality scan as /tools/ko-quality on the concatenated excerpt text. It does not download the host archive.
Featured names (KMMLU, KoBEST, and others on the buttons) appear only if that slug already has excerpts.
Use ?slug=kmmlu to open a row that is already in the list. Selecting a row writes that query so the address can be shared. Unknown slugs are ignored.
What this page does not do
- It does not host, stream, or reconstruct the original dataset. Many excerpts are format illustrations, not full items.
- A clean quality scan on a two-line excerpt is not a finding about the whole corpus.
- English names follow the same catalog display rules as dataset pages. Missing English in a source title is not filled in.