Train models to generate clearly structured advertising text for outdoor media using verifiable layout constraints across multiple display formats.

Train models to reason transparently about receivables aging, credit loss rates, and allowance movements using consumer installment finance reports.

Connect questions to policy text, supporting quotes, and coverage topics for evidence-grounded collision insurance answers.

Explore literature-based QA on endoscopic care for upper gastrointestinal bleeding, from bleeding recognition and hemostasis to risk assessment and follow-up.

Train and evaluate pharmacy-focused models with practice questions and explanations on prescriptions, drug risks, and medication safety.

Train models to generate more accurate unit tests for API endpoints using handler implementations, supporting context, and test examples.

Bring together KPI, alarm, and configuration evidence to train models to infer and justify RAN root causes from complex network observations.

Combine diverse rescue records with failure analyses to help models produce evidence-based, professional incident reviews.

Train models to assess grid requirement feasibility and changing use cases by comparing power, energy, and duration constraints across storage configurations.

Train models to assess offshore oil well intervention applicability and explain their reasoning from well facts and operational rules.

Evaluate whether a model can explain differing analog circuit conclusions by comparing source evidence and research conditions.

Analyze exactly-once semantics in distributed event processing through source code, dependencies, and the impact of code changes.

Improve a model's ability to diagnose 5G SA registration and PDU session failures with signaling clues and step-by-step explanations.

Help coding agents understand how cloud-native instrumentation generates telemetry and propagates context.

Train multi-step reasoning over plan terms, prior accumulations, and claim amounts to calculate medical cost sharing.

Trace the effects of API changes through dependency graphs and train models to identify affected objects with explainable dependency paths.

Train models to identify Arabic rhetorical terms, explain their definitions, and analyze their effects in context.

Turn real-world Korean public notices into clear learner-friendly explanations with structured details about requirements and actions.

Train models to reason about how changes to watershed nutrient sources may affect water quality, with explicit assumptions and evidence boundaries.

Train models to diagnose and repair critical job scheduler logic defects, from reproducing dependency failures to validating patches.