From Business Problems to AI Solutions: Where Does Transformation Support Fail

This collection of recent CS/SE preprints covers methods for translating business problems into ML solutions (identifying the Analytics Translation Problem and recommending research directions); scalable, formally-grounded verification of model transformations (a Cutoff Theorem for a DSLTrans fragment with Z3 implementation); pre-execution structural verification to make LLM-generated EDA code more reliable; a mutualistic neural active-learning framework for cross-project bug-report identification; PtoP, an SVGD-based seed generator that improves diversity and failure discovery for autonomous-