HHU-SyLo at SemEval-2026 Task 11: Logic in the Loop – Hybridizing LLMs and Theorem Provers for Robust Formal Reasoning

AbstractWe present our system for SemEval-2026 Task 11 on reasoning disentanglement, separating syllogistic validity from semantic plausibility. We compare direct neural inference against two neuro-symbolic pipelines: translation to first-order logic and to syllogistic triples. By offloading inference to symbolic theorem provers, these hybrid models effectively mitigate content bias and improve logical fidelity.