TUCNLP at SemEval-2026 Task 11: Neuro-Symbolic Content Stripping for Debiased Syllogistic Reasoning

Rafael Butas, Alex Lapusan, Camelia Lemnaru, Rodica Potolea


Abstract
In this paper, we present the solution submitted by TUCNLP at SemEval-2026 Task~11: Disentangling Content and Formal Reasoning in Large Language Models. The task requires predicting the formal validity of categorical syllogisms while minimizing susceptibility to content-driven biases in English and 11 additional languages. We show that a modestly-sized model (Qwen3-8B) can achieve near-perfect logical reasoning on the English validity-only subtask, and large reductions in content effect on multilingual and premise-retrieval variants, when augmented with a multi-stage neuro-symbolic pipeline: LLM-based content stripping with iterative error correction converts natural language to abstract categorical forms, a classical symbolic parser validates against the twenty-four Aristotelian syllogistic forms, and asymmetric confidence thresholds mediate between symbolic and neural decisions. Across the four subtasks (ST1 to ST4), our system achieves accuracy ranging from 91.1\% to 100\% and bias-penalized ranking scores ($\mathcal{M}$) from 31.8 to 100.0, with the main bottleneck being overconfident neural predictions that bypass symbolic verification.
Anthology ID:
2026.semeval-1.423
Volume:
Proceedings of the 20th International Workshop on Semantic Evaluation (2026)
Month:
July
Year:
2026
Address:
San Diego, California, USA
Editors:
Ekaterina Kochmar, Debanjan Ghosh, Kai North, Mamoru Komachi
Venues:
SemEval | WS
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
3408–3421
Language:
URL:
https://aclanthology.org/2026.semeval-1.423/
DOI:
Bibkey:
Cite (ACL):
Rafael Butas, Alex Lapusan, Camelia Lemnaru, and Rodica Potolea. 2026. TUCNLP at SemEval-2026 Task 11: Neuro-Symbolic Content Stripping for Debiased Syllogistic Reasoning. In Proceedings of the 20th International Workshop on Semantic Evaluation (2026), pages 3408–3421, San Diego, California, USA. Association for Computational Linguistics.
Cite (Informal):
TUCNLP at SemEval-2026 Task 11: Neuro-Symbolic Content Stripping for Debiased Syllogistic Reasoning (Butas et al., SemEval 2026)
Copy Citation:
PDF:
https://aclanthology.org/2026.semeval-1.423.pdf
Supplementarymaterial:
 2026.semeval-1.423.SupplementaryMaterial.zip