Researchers have created a large dataset called HiFi-KPI to help machines accurately extract financial Key Performance Indicators (KPIs) from company earnings reports. The dataset includes over 1.65 million paragraphs and 198k unique labels linked to the iXBRL taxonomy, which is used for public financial filings. This can aid in making more informed investment decisions by providing accurate KPIs. The researchers also released a smaller subset called HiFi-KPI-Lite, which shows that encoder-based models can achieve high accuracy in classifying and extracting KPIs. --- Why it matters: This matters to AI engineers because it provides a large-scale dataset for training and testing models that can accurately extract financial information from company reports. This can be useful for applications such as automated financial analysis and decision-making. Source: https://arxiv.org/abs/2502.15411