SKU SFT-2005 · Sold by External
SlimOrca
Product specifications
| SKU | SFT-2005 |
|---|---|
| Data type | SFT dataset |
| Volume | 517982 rows |
| Size on disk | 986 MB |
| Format | JSONL |
| Access model | PUBLIC LICENSE |
| Pricing | Free · open-source license |
| Quality score | — |
| License | MIT |
A curated ~500k-completion subset of the OpenOrca dataset released by the Open-Orca team. It keeps GPT-4-generated responses to FLAN-derived prompts and applies an additional GPT-4 pass that removes answers judged incorrect against GPT-4/human annotations, matching larger-corpus quality with far fewer rows. Rows use the ShareGPT conversations schema (system/human/gpt turns) with an optional per-turn weight.