Amazon-Llama3, Amazon-Qwen2, Amazon-Qwen-DSR1
收藏资源简介:
论文中构建了三个由大型语言模型生成的垃圾评论数据集,每个数据集包含由不同LLM生成的2500条垃圾评论。这些数据集用于研究如何检测LLM生成的垃圾评论,并评估了这些评论的欺骗性和人类相似度。数据集通过模拟欺诈者在亚马逊平台上发布垃圾评论的过程来构建,并使用GPT-4.1模型评估了评论的质量。这些数据集有助于开发更准确的垃圾评论检测方法,以应对由LLM生成的高级垃圾评论带来的挑战。
This paper constructs three spam review datasets generated by large language models (LLMs). Each dataset contains 2500 spam reviews produced by distinct LLMs. These datasets are utilized for researching the detection of LLM-generated spam reviews, and to evaluate the deceptiveness and human-likeness of such reviews. The datasets are built by simulating the process of fraudsters posting spam reviews on the Amazon marketplace, and the quality of the generated reviews is assessed using the GPT-4.1 model. These datasets facilitate the development of more accurate spam review detection methods to address the challenges posed by advanced LLM-generated spam reviews.

- 1Detecting LLM-Generated Spam Reviews by Integrating Language Model Embeddings and Graph Neural Network清华大学 · 2025年



