This paper presents our approach for SemEval-2026 Task 11: Disentangling Content and Formal Reasoning in Large Language Models. We propose a neuro-symbolic teacher-student framework that utilizes DeepSeek-R1 as a Logical Auditor to generate a high-fidelity training corpus. We distill these analytical behaviors into Qwen-3 models using Low Rank Adaptation (LoRA), focusing on teaching the mechanics of logic rather than simple label matching. Our system yields robust results across both subtasks, with a ranking score of 39.81 (96.86% accuracy) on Subtask 1 and 26.02 on Subtask 3. However, validity bias partially persists, so we conclude that while structured distillation substantially mitigates belief bias, fully disentangling logical validity from plausibility remains a central challenge for future development.

Claims about the grammatical competence of multilingual language models vary sharply with how competence is measured, yet the interaction between evaluation paradigm, post-training, and language resource availability has not been systematically examined. We evaluate base and post-trained models from six families on MultiBLiMP, a syntactic minimal-pair benchmark covering 101 languages, using four evaluation methods. We report three principal findings. First, post-training degrades grammatical competence, but the magnitude of this effect is determined by post-training recipe rather than model scale, and scaling reduces it unevenly, with low-resource languages bearing the highest cost. Second, post-trained models retain grammatical knowledge they cannot articulate through explicit prompting, yet this is measurable only in high-resource languages, because near-chance baselines in low-resource settings leave little knowledge to hide. Third, native-language prompting recovers otherwise hidden competence on low-resource languages, demonstrating that only high-resource languages can be probed directly from unprompted probabilities. We conclude that multilingual grammatical evaluation must adopt language-informed, multi-paradigm protocols to avoid systematically underestimating low-resource abilities.