Large language models (LLMs) suffer from a long-tail deficit: culturally specific facts, particularly those concerning underrepresented regions such as Latin…
机构:巴黎综合理工
来源:arXiv 2609.18317 | AI4Papers 论文推荐平台