交易基础模型用于金融交易数据RAG：与文本嵌入的对比评估
原标题：Transaction Foundation Models for Retrieval-Augmented Generation over Financial Transaction Data: A Comparative Evaluation with Text Embeddings
AI 摘要
该论文在IBM合成反洗钱交易数据集上构建了包含1,026个问题、带检索级金标文档标注的确定性问答基准，用于评估面向金融交易数据的RAG。作者将NVIDIA 29M参数的Transaction Foundation Model（TFM）与五个通用及金融领域文本嵌入模型在规模和数据匹配条件下对比，发现TFM在Recall@5（0.81 vs 0.25–0.39）和端到端答案准确率（66.7% vs 27.5%）上显著领先。但用通用句编码器嵌入人工设计的聚合摘要文本的简单流程整体上可匹敌甚至超过TFM，并在银行、货币、支付格式等低基数实体类别上占优，说明交易原生基础模型的优势具有类别依赖性。作者公开了基准生成与评估代码。
正文节选
*1 Transaction Foundation Models for Retrieval-Augmented Generation over Financial Transaction Data: A Comparative Evaluation with Text Embeddings Abstract Domain-specific tabular and transaction foundation models are increasingly proposed as drop-in representation learners for financial applications, but their value relative to off-the-shelf text embedding models for retrieval-augmented generation (RAG) over structured data remains underexamined. We construct a deterministic, template-generated