LLM Pipeline for Logical Understanding of Enthymemes
Researchers propose a pipeline to systematically decode enthymemes—arguments with implicit premises—by integrating large language models (LLMs) and a neuro-symbolic reasoner. The method uses one LLM to generate missing premises from explicit text, another to translate natural language into logical formulas, and a reasoner to verify entailment. This addresses the gap between NLP methods that identify enthymemes but ignore logic, and logic-based approaches that require pre-existing knowledge bases. The work is published on arXiv (2603.06114v2) and aims to enable automated logical analysis of real-world arguments.
Key facts
- arXiv paper 2603.06114v2 proposes a pipeline for enthymeme decoding
- Pipeline uses two LLMs: one for premise generation, one for logical translation
- Neuro-symbolic reasoner checks logical entailment
- Addresses lack of systematic method for translating enthymemes into logical arguments
- Real-world arguments in text are often enthymemes with implicit premises
- NLP methods identify enthymemes but do not decode underlying logic
- Logic-based approaches require sufficient knowledge base formulae
- Method integrates LLMs with neuro-symbolic reasoning
Entities
Institutions
- arXiv