pszemraj/

infinity-instruct-7m-T2T_en

Overview of the filters and processing steps applied to the original dataset:

Conversation Length Filter: Filtered out conversations longer than 5 exchanges Language Detection Filter: Removed non-English conversations using the default language col Instruction-Response Extraction: Extracted the first human instruction and the corresponding LLM response from each conversation. Refusal Response Filter: Removed rows w/ responses containing… See the full description on the dataset page: https://huggingface.co/datasets/pszemraj/infinity-instruct-7m-T2T_en.