- Fix FG_82 reference filename to match JSONL (e1ff48f70a9c9e1d035ca5dd7064e439e8d4d71a)
Co-authored-by: Sun Yixuan <tsingsx@users.noreply.huggingface.co>
Previously only the 3 Chinese jsonls were declared as configs. The 3 English
jsonls in *_en/ were left to viewer auto-discovery, which can cause schema
collisions and 'dataset viewer should be available soon' failures. Declare
all 6 splits with explicit config names.
English version of file_generation dataset.
- jsonl text fields (question/reference/rubrics) translated
- binary file contents (xlsx/csv/docx headers + cell values) translated
- file names + sheet names made self-consistent across jsonl and on-disk files
- output_type mapped to English
- Real data cell values (currency units, search-keyword examples, code labels) kept in Chinese so evaluation against the .xlsx data still works.
Add data_visualization_en: full English localization (jsonl + binary files)
English version of data_visualization dataset.
- jsonl text fields (question/reference/rubrics) translated
- binary file contents (xlsx/csv/docx headers + cell values) translated
- file names + sheet names made self-consistent across jsonl and on-disk files
- output_type mapped to English
- Real data cell values (currency units, search-keyword examples, code labels) kept in Chinese so evaluation against the .xlsx data still works.
Add QA_en: full English localization (jsonl + binary files)
English version of QA dataset.
- jsonl text fields (question/reference/rubrics) translated
- binary file contents (xlsx/csv/docx headers + cell values) translated
- file names + sheet names made self-consistent across jsonl and on-disk files
- output_type mapped to English
- Real data cell values (currency units, search-keyword examples, code labels) kept in Chinese so evaluation against the .xlsx data still works.
