Turn road signs, lane restrictions, vehicle classes, and time conditions into queryable rule relations for Robotaxi maps.

Evaluate whether models can accurately extract obligations, pricing conditions, payment timelines, and exceptions from procurement contracts.

Turn out-of-home advertising compliance requirements into structured constraints for precise extraction of brand rules, disclosures, and placement restrictions.

Extract component-to-material relationships from footwear product descriptions with source evidence for traceable, structured catalog data.

Train models to turn natural-language process analysis requests into clear, structured specifications.

Extract key entities such as devices, interfaces, and protocols from network configuration change records to build models that better understand network operations.

Extract access, booking, fee, and equipment-use rules from microscopy facility webpages to help models interpret operational policies accurately.

Train models to understand luxury hotel packages and resort experiences and extract their key attributes.

Evaluate how well models extract strawberry growth stages and phenological attributes from cultivation webpages, with supporting evidence spans.

Turn experiment protocols into structured information about reagents, equipment, steps, and conditions for virtual laboratory workflows.

Evaluate how accurately models identify flow sensor models and link key specifications across product pages, selection guides, and technical tables.

The Clinic Electronic Medical Record Q&A Dataset focuses on image data mining in the healthcare field.

Identify parties, courts, businesses, and contracts in commercial litigation documents with annotations tailored to the language of legal proceedings.

Pair commercial construction opportunity notices with structured JSON to train models for accurate field extraction and schema-compliant output.

Extract power-device specifications from manufacturer product descriptions while preserving the conditions behind key parameters.

Train models to identify people, places, events, and textual evidence accurately using Korean passages with detailed JSON annotations.

Extract operating conditions, contact states, and related sensor-state information from temperature sensor web text for focused domain evaluation.

Evaluate domain-specific information extraction across a broad range of common flow sensor operating states.

Convert virtual laboratory equipment descriptions into structured JSON covering components, parameters, connections, and applicable steps.

Extract key provisions, effectiveness conditions, and financing implications from complex leveraged loan amendments and waivers.