Neurosymbolic Language Reasoning as Satisfiability Modulo Theory
arXiv:2602.18095v1 Announce Type: new Abstract: Natural language understanding requires interleaving textual and logical reasoning, yet large language models often fail to perform such reasoning reliably. Existing neurosymbolic systems combine LLMs with solvers but remain limited to fully formalizable tasks such as math or program synthesis, leaving natural documents with only partial logical structure unaddressed. We introduce Logitext, a neurosymbolic language that represents documents as natural language text constraints (NLTCs), making partial logical structure explicit. We develop an algorithm that integrates LLM-based constraint evaluation with satisfiability modulo theory (SMT) solving, enabling joint textual-logical reasoning. Experiments on a new content moderation benchmark, together with LegalBench and Super-Natural Instructions, show that Logitext improves both accuracy and coverage. This work is the first that treats LLM-based reasoning as an SMT theory, extending neurosy
arXiv:2602.18095v1 Announce Type: new Abstract: Natural language understanding requires interleaving textual and logical reasoning, yet large language models often fail to perform such reasoning reliably. Existing neurosymbolic systems combine LLMs with solvers but remain limited to fully formalizable tasks such as math or program synthesis, leaving natural documents with only partial logical structure unaddressed. We introduce Logitext, a neurosymbolic language that represents documents as natural language text constraints (NLTCs), making partial logical structure explicit. We develop an algorithm that integrates LLM-based constraint evaluation with satisfiability modulo theory (SMT) solving, enabling joint textual-logical reasoning. Experiments on a new content moderation benchmark, together with LegalBench and Super-Natural Instructions, show that Logitext improves both accuracy and coverage. This work is the first that treats LLM-based reasoning as an SMT theory, extending neurosymbolic methods beyond fully formalizable domains.
Executive Summary
This article introduces Logitext, a neurosymbolic language that enables joint textual-logical reasoning by representing documents as natural language text constraints. The algorithm integrates LLM-based constraint evaluation with satisfiability modulo theory solving, improving accuracy and coverage in content moderation, legal, and instruction-based tasks. Logitext extends neurosymbolic methods beyond fully formalizable domains, treating LLM-based reasoning as an SMT theory. The approach addresses the limitations of existing neurosymbolic systems and large language models in performing reliable textual and logical reasoning.
Key Points
- ▸ Introduction of Logitext, a neurosymbolic language for joint textual-logical reasoning
- ▸ Integration of LLM-based constraint evaluation with satisfiability modulo theory solving
- ▸ Extension of neurosymbolic methods beyond fully formalizable domains
Merits
Improved Accuracy and Coverage
Logitext improves both accuracy and coverage in various tasks, including content moderation, legal, and instruction-based tasks.
Demerits
Limited Generalizability
The approach may be limited to specific domains or tasks, and its generalizability to other areas remains to be explored.
Expert Commentary
The introduction of Logitext represents a significant advancement in neurosymbolic language reasoning, enabling more accurate and reliable textual-logical reasoning. The integration of LLM-based constraint evaluation with SMT solving is a novel approach that addresses the limitations of existing neurosymbolic systems. However, further research is needed to explore the generalizability of this approach to other domains and tasks, and to address potential concerns around explainability and transparency. The implications of this work are far-reaching, with potential applications in various fields, including content moderation, law, and education.
Recommendations
- ✓ Further research on the generalizability of Logitext to other domains and tasks
- ✓ Exploration of the potential applications of Logitext in various fields, including content moderation, law, and education