Train models to generate energy optimization code from flow battery scheduling requirements and operating constraints.

Train models to answer questions about decorative arts materials and techniques with clear evidence from collection sources.

Train models to make verifiable patent novelty assessments grounded in single-document prior art and claim-feature mapping.

Pairing exam syllabus pages with evidence-grounded candidate Q&A, this dataset helps models answer questions about tested content and exam rules.

Train models to detect conflicts across sources and judge whether citations adequately support synthetic social advertising claims.

Evaluate whether models can detect conflicts, duplicates, and incompatible arrangements across orders in pediatric inpatient care plans.

Make SAT answer choice decisions more retrievable and explainable with evidence that supports correct answers and refutes distractors.

Train models to handle major appliance shopping requests more accurately with complete search and comparison tool-call trajectories.

Retrieve the case file passages behind summary judgment assertions and strengthen evidence-grounded legal search.

Evaluate whether models can detect ambiguity in question bank text, judge answerability, and explain their conclusions using supporting evidence.

Train models to understand German grammar in context with instructional web documents, rule judgments, and cross-paragraph QA supervision.

Evaluate whether models can detect omissions, ordering conflicts, and unmet prerequisites in lung resection workflows using structured steps and constraint annotations.

Connect venue specifications from business hotel pages with source evidence for more reliable meeting and event information retrieval.

Train models to diagnose FTTB faults more reliably by grounding conclusions in network dependencies and operational evidence.

Train models to identify concurrency mechanisms, shared state, and potential race conditions in distributed cache code through source-based question answering.

Retrieve key employment termination, handover, and severance clauses with supporting excerpts from source legal documents.

Train models to identify why simulation configurations fail and assess their potential effects using JSON examples with conflict locations and scientific rationale.

Examine the boundaries of licensing and distribution rights through linked claims, contract evidence, and verification outcomes.

Retrieve compensation and benefits clauses with supporting contract evidence to answer employment agreement questions more precisely.

Retrieve scenario-linked evidence for confidentiality exceptions and support precise proof analysis and boundary assessment.