Selecting high-impact data is essential for enhancing model performance and informing downstream machine learning workflows. A new feature allows users to group and investigate specific data slices in Model, enabling analysis of data rows based on metadata values, which helps prioritize and save high-impact data for improved data quality and debugging. Slices in Model offer dynamic, real-time views of data rows that match specific search criteria, and auto-generated slices facilitate trend analysis and model evaluation. Users can also filter data rows on metadata values for targeted performance analysis. An improved data export process is in beta, offering greater control over export fields and the ability to export specific data sets from projects. This new export functionality supports the integration of AI workflows with adjacent tools, allowing users to include or exclude relevant attributes during export, and is accessible through the UI or Python SDK.