研究人员正在探索语言模型传统指令微调的替代方案，特别是针对小型和多语言模型。一篇论文研究了上下文学习（ICL）在非英语语言和不同模型规模下指令遵循方面的有效性，发现ICL在此类场景下的性能有所下降。另一项研究引入了M-DaQ，一个用于创建高质量、多样化多语言指令微调数据集的框架，该框架能提升模型在18种语言上的性能。第三篇论文提出了一种名为加权上下文影响（wICI）的数据选择方法，用于识别有效的指令微调数据，在数据受限的情况下优于现有基线。 AI
影响 新的多语言指令微调和数据选择方法可以提高LLM在不同语言上的性能和可访问性。
排序理由 该集群包含多篇arXiv论文，详细介绍了语言模型指令微调和数据选择方面的新研究。
AI 生成摘要 · Google Gemini · 来自 4 个来源。 我们如何撰写摘要 →
arXiv:2503.01611v3 Announce Type: replace Abstract: Instruction following is a critical ability for Large Language Models to perform downstream tasks. The standard approach to instruction tuning has relied on a specific phase of supervised fine-tuning over curated instruction dat…
arXiv:2509.15549v2 Announce Type: replace Abstract: Multilingual instruction fine-tuning (IFT) empowers large language models to generalize across diverse linguistic and cultural contexts; however, high-quality, systematically curated multilingual IFT datasets remain scarce. To a…
arXiv:2604.25132v1 Announce Type: new Abstract: Instruction-tuning datasets often contain substantial redundancy and low-quality samples, necessitating effective data selection methods. We propose an instruction data selection framework based on weighted in-context influence (wIC…
Instruction-tuning datasets often contain substantial redundancy and low-quality samples, necessitating effective data selection methods. We propose an instruction data selection framework based on weighted in-context influence (wICI), which measures how effectively each candidat…