calibrate_qa_mapper

Calibrating QA pairs requires aligning them with reference text to ensure matching language style and context. This process automates data transformation by adjusting both questions and answers to conform to your desired reference styles, enhancing overall dataset accuracy.

QA calibration for reference text alignment is a data preprocessing technique that ensures question-answer pairs match the language style and context of a given reference text. It processes and adjusts questions and answers to conform to specific reference styles, improving accuracy in Q&A datasets.

Yes, data_juicer is required to process samples for language style adjustment. The calibration process relies on data_juicer for sample processing and may also require external LLM API calls to accurately align question-answer pairs with the reference text.

To refine FAQs based on a particular language style or context, use a calibration skill that aligns question-answer pairs with reference text. Specify your input and output JSON file paths along with an LLM model to automate the data transformation.

Yes, QA calibration potentially requires external LLM API calls to adjust Q&A datasets. While data_juicer handles the core sample processing, external LLM APIs are utilized to ensure accurate language style adjustment and reference text alignment.