Distilling Answer-Set Programming Rules from LLMs for Neurosymbolic Visual Question Answering

In three linesMethod to distill Answer-Set Programming (ASP) rules from LLMs for Visual Question Answering (VQA). The approach uses VQA dataset examples to guide the LLM in extending an initial reasoning theory, with validation by the ASP solver. Demonstrates effectiveness across multiple datasets with few examples.

Summary generated by Claude — human-verified