An Agentic Framework for Neuro-Symbolic Programming

TL;DR

This paper introduces AgenticDomiKnowS (ADS), a framework that automates the creation of neuro-symbolic programs from free-form descriptions, significantly reducing development time and supporting human-in-the-loop refinement.

Contribution

ADS automates the translation of natural language descriptions into DomiKnowS programs, eliminating the need for expert knowledge of the library's syntax.

Findings

Reduces neuro-symbolic programming development time from hours to 10-15 minutes.

Supports human-in-the-loop refinement for improved accuracy.

Enables both experienced and new users to rapidly construct programs.

Abstract

Integrating symbolic constraints into deep learning models could make them more robust, interpretable, and data-efficient. Still, it remains a time-consuming and challenging task. Existing frameworks like DomiKnowS help this integration by providing a high-level declarative programming interface, but they still assume the user is proficient with the library's specific syntax. We propose AgenticDomiKnowS (ADS) to eliminate this dependency. ADS translates free-form task descriptions into a complete DomiKnowS program using an agentic workflow that creates and tests each DomiKnowS component separately. The workflow supports optional human-in-the-loop intervention, enabling users familiar with DomiKnowS to refine intermediate outputs. We show how ADS enables experienced DomiKnowS users and non-users to rapidly construct neuro-symbolic programs, reducing development time from hours to 10-15 minutes.