用新方法选训练数据，效果更好且更省资源
RICo: Refined In-Context Contribution for Automatic Instruction-Tuning Data Selection
- 基于上下文学习量化样本贡献，精细筛选数据
- 仅用15%数据就超越全量数据，提升5.42个百分点
- 适合追求高效训练的模型开发者
指令微调中的数据选择对提升大语言模型性能、降低训练成本至关重要。本文提出一种无梯度的精细化上下文贡献度量方法RICo，可量化单个样本在任务级与全局级模型表现中的细粒度贡献。RICo能更精准识别高贡献数据，从而优化指令微调效果。我们进一步设计一种轻量级选择范式，基于RICo得分实现线性复杂度的可扩展数据筛选。在三个LLM、12个基准测试和5组成对评估中广泛验证了有效性。显著地，在LLaMA3.1-8B上，仅使用15%的RICo选中数据即比全量数据提升5.42个百分点，并优于主流选择方法2.06个百分点。进一步分析显示，RICo选出的高贡献样本兼具任务多样性与适中难度，而非仅最难题目。
原文摘要 · Abstract (English)
Data selection for instruction tuning is crucial for improving the performance of large language models (LLMs) while reducing training costs. In this paper, we propose Refined Contribution Measurement with In-Context Learning (RICo), a novel gradient-free method that quantifies the fine-grained contribution of individual samples to both task-level and global-level model performance. RICo enables more accurate identification of high-contribution data, leading to better instruction tuning. We further introduce a lightweight selection paradigm trained on RICo scores, enabling scalable data selection with a strictly linear inference complexity. Extensive experiments on three LLMs across 12 benchmarks and 5 pairwise evaluation sets demonstrate the effectiveness of RICo. Remarkably, on LLaMA3.1-8B, models trained on 15% of RICo-selected data outperform full datasets by 5.42% points and exceed the best performance of widely used selection methods by 2.06% points. We further analyze high-contribution samples selected by RICo, which show both diverse tasks and appropriate difficulty levels, rather than just the hardest ones.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。