Preference data for training and aligning Saraiki language models using Direct Preference Optimization (DPO). Saraiki (سرائیکی) is an Indo-Aryan language spoken primarily in Pakistan, particularly across southern Punjab and neighboring regions.
Dataset Card
Preference data for training and aligning Saraiki language models using Direct Preference Optimization (DPO). Saraiki (سرائیکی) is an Indo-Aryan language spoken primarily in Pakistan, particularly across southern Punjab and neighboring regions. It has a rich linguistic, literary, and cultural heritage, with its own vocabulary, grammatical structures, and regional varieties. Despite its significant speaker community, Saraiki remains a low-resource language in Natural Language Processing (NLP), with relatively limited digital datasets, language models, and computational resources compared with widely supported languages. The Saraiki Direct Preference Optimization Dataset is a curated…
Excerpt from the card by Muhammad Farjad Ali Raza, licensed cc-by-nc-4.0.
Details
- Repository
- themohal/saraiki-dpo-sft-dataset
- Publisher
- Muhammad Farjad Ali Raza
- Task category
- Not stated by the source
- Tags
- saraiki, jataki-saraiki, dpo
- Size category
- Not stated by the source
- Languages
- skr
- Revision
- 28e30a4a6bd8059e560ec4079164948fa2669410
- Last updated
- 2026-10-11
Files
3 files, 412.1 KB in total.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| data/train-00000-of-00001.parquet | Data | 403.9 KB | — |
| README.md | Documentation | 5.7 KB | — |
| .gitattributes | Repository | 2.5 KB | — |
License and Download
- License
- cc-by-nc-4.0
- Access
- Access requested at publisher
Muhammad Farjad Ali Raza grants access through its official repository on Hugging Face. Read the license.