1 paper
Mert Ünsal, Timon Gehr, Martin Vechev
We present the first correct-by-construction learning-based system for step-by-step mathematical integration. The key idea is to learn a policy, represented by a GPT transformer mo…