japan-ai-official/edinet-bench-earnings-forecast-curated

这是一个来自EDINET-Bench欺诈检测数据集的高质量子集，包含了39个经过精心挑选的样本，这些样本通过高级难度评估和模型评估技术进行筛选。数据集的选取基于难度分数、一致性分数、模型协议和响应质量。适用于欺诈检测模型的评估和基准测试，特别是在具有挑战性的金融案例上；用于训练模型以识别困难的金融欺诈任务；用于研究金融欺诈检测能力；以及用于教育目的，了解模型在欺诈检测中的局限性。

This is a high-quality subset of the EDINET-Bench fraud detection dataset, containing 39 carefully selected samples curated using advanced difficulty assessment and model evaluation techniques. The selection is based on the difficulty score, consistency score, model agreement, and response quality. It is suitable for fraud detection model evaluation and benchmarking on challenging financial cases, training models on difficult financial fraud identification tasks, researching financial fraud detection capabilities, and educational purposes to understand model limitations in fraud detection.