UFAL-CUNI at SemEval-2026 Task 11: An Efficient Modular Neuro-symbolic Method for Syllogistic Reasoning

TL;DR

This paper introduces a modular neuro-symbolic system combining small LLMs and symbolic reasoning for syllogistic reasoning, achieving competitive accuracy and outperforming zero-shot baselines.

Contribution

The paper presents an efficient neuro-symbolic approach that integrates symbolic provers with small LLMs for reasoning tasks, demonstrating improved performance over baseline methods.

Findings

System achieves competitive accuracy on syllogistic reasoning tasks.

Outperforms zero-shot LLM baselines in the same parameter range.

Limited multilingual capabilities observed in small LLMs.

Abstract

This paper describes our system submitted to SemEval-2026 Task 11: Disentangling Content and Formal Reasoning in Large Language Models. We present an efficient modular neuro-symbolic approach, combining a symbolic prover with small reasoning LLMs (4B parameters). The system consists of an LLM-based parser that translates natural language syllogisms to a first-order logic (FOL) representation, an automated theorem prover, and two optional modules: machine translation for multilingual inputs and a symbolic retrieval component for the identification of relevant premises. The system achieves competitive accuracy and relatively low content effect on most subtasks. Our ablations show that this approach outperforms LLM-based zero-shot baselines in this parameter size range, but also reveal limited multilingual capabilities of small LLMs. Finally, we include a discussion of the task's main ranking metric and analyze its limitations.