Train models to interpret patent license scope, commercial terms, and legal obligations through contract-grounded questions and answers.

Train models to determine whether power outage emergency orders substantiate claims using document excerpts, verification labels, and evidence-location annotations.

Train and evaluate weightlifting process verifiers with step-level constraint and plan consistency labels.

Train models to calculate patient responsibility and annual out-of-pocket status from plan rules and sequential claim histories.

Help models answer specialized patent portfolio questions accurately with document-grounded answers and traceable evidence.

Follow the complete late-night ordering workflow from understanding a request to submitting an order, and train models to orchestrate food delivery tasks reliably.

Evaluate fact verification on employment contract terms and termination clauses using source text and evidence-based assessments.

This dataset is used to identify students' operational behavior in an experimental teaching environment.

Help models make evidence-based coverage determinations by focusing on easily confused boundaries between collision coverage and adjacent protections.

Retrieve key delivery and acceptance clauses with evidence to support accurate RAG search across performance documents.

Evaluate how models handle IT incident deadlines and escalation rules with structured SLA cases and verifiable rewards.

Help models clarify misconceptions with textbook-based explanations and strengthen specialized concept reasoning.

This dataset provides video trajectories of cleaning robots in various environments to support intelligent navigation and planning research.

Train models to assess biobank cohort representativeness and selection bias using evidence from multiple papers without overstating what the evidence supports.

Train models to derive indoor DAS topologies and explain coverage design decisions through structured problems and constraint analysis.

Train models to calculate RF link budgets and assess feasibility from typical indoor DAS conditions.

Improve long-contract analysis with challenging examples of defined-term scope, nested conditions, and cross-references that local context can miss.

Connect SAT editing questions with applicable rules and answer evidence to build grounded tutoring and explanation systems.

Connect source text with supporting passages on metalworking heritage to help models retrieve precise evidence for technique questions.

Connects license scope questions to relevant assessments and contract evidence for training and evaluating legal document retrieval models.