rule-distillation

This paper presents a method for distilling answer-set programming rules from large language models to enhance neurosymbolic visual question answering, showing that only a few examples are needed to generate correct rules.