From Reweighting to Rewriting: Unlocking the Intervention Effects of Influential Samples in Training Data Attribution

• 类型：arxiv
• 标识：2609.02771
• 链接：https://arxiv.org/abs/2609.02771
• 主分类：engineering
• 形态：method
• TLDR：Training data attribution (TDA) aims to identify training examples that shape model behavior, but its intervention value depends on both which examples are selected and how they are modified. Influence functions (IF) estimate behavioral changes under infinitesimal reweighting, yet IF-selected examples often show limited advantages over random selection under conventional weight-based interventions. This raises the question of whether influential examples lack intervention value or whether reweighting fails to realize their behavioral leverage.We introduce influence-guided response rewriting, w
• 待LLM分类：否
• 标题中文：从重加权到改写：解锁训练数据归因中有影响力样本的干预效果
• TLDR中文：训练数据归因（TDA）旨在识别塑造模型行为的训练样本，但其干预价值既取决于样本选择，也取决于修改方式。影响函数（IF）估计的是无穷小重加权下的行为变化，但在基于权重的常规干预下，IF 选出的样本相对随机选择往往优势有限。这引出问题：有影响力样本缺乏干预价值，还是重加权未能释放其行为杠杆？我们提出影响引导的响应改写方法……
• 来源文件：
• /inbox/tom/_candidates/2026-09-11-agent-rag-longcontext-candidates.json