Inductive Program Synthesis via Iterative Forward-Backward Abstract Interpretation

A key challenge in example-based program synthesis is the gigantic search space of programs. To address this challenge, various work proposed to use abstract interpretation prune space. However, most existing approaches have focused only on forward interpretation, and thus cannot fully exploit power interpretation. In paper, we propose a novel approach inductive via iterative forward-backward The computes possible outputs given inputs, while backward inputs outputs. By iteratively performing two interpretations an alternating fashion, can effectively determine if any completion each partial as candidate satisfy input-output examples. We apply our standard formulation, syntax-guided (SyGuS), thereby supporting wide range tasks. implemented evaluated it set benchmarks from prior work. experimental results show that significantly outperforms state-of-the-art thanks sophisticated techniques.