贪心信息投影用于大语言模型数据选择
Aligning the Query Space: Greedy Information Projection for Language Model Data Selection
AI总结：
本文提出贪心信息投影框架，通过最大化子集与任务查询信号的互信息，平衡质量与多样性，高效选择数据集进行微调。
摘要：
我们提出了贪心信息投影（GIP），一种 principled 框架，用于选择大型语言模型微调的训练示例。GIP 将选择过程定义为最大化子集与任务特定查询信号之间的互信息，这些信号可能来自 LLM 质量判断、元数据或其他来源。该框架涉及优化由数据和查询嵌入定义的闭式互信息目标，自然平衡质量与多样性。优化此分数等同于最大化查询嵌入矩阵在所选数据跨度上的投影，这为质量和多样性的同时出现提供了几何解释。基于此观点，我们采用快速贪心匹配追踪程序，具有高效的投影更新。在遵循指令和数学推理数据集上，GIP 选择小的子集，匹配完整数据微调，同时仅使用少量示例和计算，统一了质量意识和多样性意识的选择，以实现高效的微调。
英文摘要：
Data selection for language models is often framed as balancing example quality and diversity. We argue that both are consequences of a more fundamental principle: selected examples should preserve the downstream query space induced by instructions, task signals, or retrieval needs. We present Greedy Information Projection (GIP), a query-aligned method that requires only candidate embeddings and a score signal from LLM judgments, metadata, or intrinsic geometry. GIP greedily selects examples whose embedding span explains the largest residual component of task/query scores, yielding a fast matching-pursuit selector. A Gaussian projection view connects this update to maximizing mutual information, equivalently minimizing the residual volume of the query subspace left unexplained by selected data; this explains how quality and diversity emerge from one objective. Empirically, GIP matches or surpasses full-data fine-tuning using small instruction and reasoning subsets, while pretraining and RAG passage-selection experiments show that the same residual principle transfers beyond supervised fine-tuning.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
