crossingminds/credit_card_3k

Credit Card 3k 数据集是一个合成任务数据集，用于评估大型语言模型（LLM）的域自适应能力，特别是比较上下文学习与微调的效果。该数据集包含3,111对信用卡交易字符串及其对应的商家名称。这些字符串对于人类来说很容易理解，但对于机器来说却难以处理，因为它们通常包含多种代码、数字、缩写以及有时混淆的支付处理器名称（如PayPal或SQ）。数据集可用于上下文学习（少量示例）以及用于微调LLM。

The Credit Card 3k Dataset is a synthetic task dataset created to evaluate the domain adaptation of Large Language Models (LLMs), specifically to compare the efficacy of in-context learning versus fine-tuning. The dataset consists of 3,111 pairs of credit card transaction strings and their corresponding merchant names. These strings are trivial for humans to understand but challenging for machines due to the inclusion of various codes, numbers, abbreviations, and sometimes confounding payment processor names like PayPal or SQ. The dataset can be used for in-context learning (few-shot examples) as well as for fine-tuning LLMs.