We present our submission to SemEval-2026 Task 7, focusing on the MCQ track, where models must identify culturally specific answers across language-region locales. Our system augments a compact open-source model with locale-targeted web retrieval at inference time, requiring no task-specific fine-tuning, and places 10th on the leaderboard. Beyond the submitted system, we explore how retrieval depth and search localization affect performance across locales, finding that localizing search parameters meaningfully shifts the geographic composition of retrieved sources and that gains from retrieval are most pronounced for lower-resource locales. We also investigate whether culturally informed prompt framing can complement retrieval, finding that it does, but only when grounding context is present. Taken together, our results point to inference-time web grounding as a practical path toward more culturally aware NLP under resource constraints.

This paper describes MIDAS_SYNUR, a system developed for the MEDIQA-SYNUR task at ClinicalNLP 2026 on observation extraction from nurse dictations. The primary system adopts a single-prompt, field-rich few-shot strategy using GPT-5.2, jointly generating all schema fields in one structured output. Few-shot demonstrations are curated and grouped by value type, with five examples per type, promoting consistency across heterogeneous value distributions while leveraging global context to resolve cross-field dependencies. To analyze design trade-offs, this holistic strategy is compared against a field-wise decomposed prompting baseline, where each schema field is extracted independently using explicit positive and NULL demonstrations to improve absence detection and reduce cross-field interference. Zero-shot variants of both approaches are also evaluated to isolate the contribution of in-context examples. The results highlight inference-time prompting as a simple, reproducible, and competitive baseline for large-scale clinical observation extraction from conversational nurse dictations. Keywords: Prompt Engineering, Few-Shot Prompting, In-Context Learning, Structured Output