An Augmented MetiTarski Dataset for Real Quantifier Elimination Using Machine Learning.

We contribute a new dataset composed of more than 41K MetiTarski challenges that can be used to investigate applications of machine learning (ML) in determining efficient variable orderings in Cylindrical Algebraic Decomposition. The proposed dataset aims to address inadvertent bias issues present in prior benchmarks, paving the way to development of robust, easy-to-generalize ML models.