CONF RamirezContreras_EMNLP_2026/IDIAP Plausible but Wrong: A Statistical Evaluation of GPT Models on High-Stakes Migration Statistics Ramirez Contreras, Hector Manuel Annapureddy, Ravinithesh Gatica-Perez, Daniel data proxies GPTs Large Language Models LLM migration numerical hallucination open data web-based AI systems EXTERNAL https://publications.idiap.ch/attachments/papers/2026/RamirezContreras_EMNLP_2026.pdf PUBLIC Proceedings of the First Workshop on Pluralistic AI & NLP: Diversity-aware, Sociotechnical, Responsible Alignment (PANDORA) 2026