ITLC Achieves Top-5 Ranking at SemEval-2026 with Novel LLM Reasoning Method
A research team from ITLC has introduced a novel method to enhance formal reasoning in Large Language Models (LLMs), specifically addressing content effects and biases in multi-lingual contexts. Published on arXiv, the study details an approach that utilizes explicit structural abstraction to transform syllogisms into canonical logical representations. By applying deterministic parsing to determine validity, the method significantly reduces biases without requiring complex fine-tuning or activation-level interventions. The technique was evaluated on the SemEval-2026 Task 11 multilingual benchmark, where it achieved top-5 rankings across all subtasks. This performance highlights the method as a competitive alternative to existing resource-intensive strategies for improving LLM reasoning capabilities. The paper, titled "ITLC at SemEval-2026 Task 11: Normalization and Deterministic Parsing for Formal Reasoning in LLMs," was submitted by authors including Wicaksono Leksono Muhamad and Samuel Cahyawijaya. The findings contribute to the fields of Computation and Language and Artificial Intelligence, offering a streamlined solution for enhancing logical consistency in AI systems across diverse languages.
Wire timeline
ITLC Achieves Top-5 Ranking at SemEval-2026 with Novel LLM Reasoning Method
A research team from ITLC has introduced a novel method to enhance formal reasoning in Large Language Models (LLMs), specifically addressing content effects and biases in multi-lingual contexts. Published on arXiv, the study details an approach that utilizes explicit structural abstraction to transform syllogisms into canonical logical representations. By applying deterministic parsing to determine validity, the method significantly reduces biases without requiring complex fine-tuning or activation-level interventions. The technique was evaluated on the SemEval-2026 Task 11 multilingual benchmark, where it achieved top-5 rankings across all subtasks. This performance highlights the method as a competitive alternative to existing resource-intensive strategies for improving LLM reasoning capabilities. The paper, titled "ITLC at SemEval-2026 Task 11: Normalization and Deterministic Parsing for Formal Reasoning in LLMs," was submitted by authors including Wicaksono Leksono Muhamad and Samuel Cahyawijaya. The findings contribute to the fields of Computation and Language and Artificial Intelligence, offering a streamlined solution for enhancing logical consistency in AI systems across diverse languages.
cs.AI updates on arXiv.org