Neuro-Symbolic Framework Bridges Natural Language Reasoning with NARS Formal Logic

A new neuro-symbolic framework addresses limitations in large language models by translating natural-language reasoning problems into executable formal representations. The approach uses first-order logic and Narsese, the language of the Non-Axiomatic Reasoning System. To support this direction, researchers introduced NARS-Reasoning-v0.1, a benchmark containing natural-language reasoning problems paired with FOL forms and executable Narsese programs. Each problem includes three gold labels: True, False, and Uncertain. A deterministic compilation pipeline was developed to convert FOL to executable Narsese. Validation occurs through runtime execution in OpenNARS for Applications, ensuring symbolic targets are both syntactically correct and behaviorally aligned with intended reasoning. This work aims to enhance reliability in tasks requiring explicit symbolic structure, multi-step inference, and interpretable uncertainty. The paper was announced as arXiv:2604.18873v1.