unlearning-cleanslate/generations-llama-3_1-8b-pre_val

该数据集是一个多任务推理评估集合，包含ARC挑战赛和多个BBH（Big-Bench Hard）任务，如布尔表达式、因果判断、日期理解、歧义消解、形式谬误、几何形状、逻辑推理、电影推荐、多步算术、导航、对象计数、表格数据推理和颜色对象推理等。每个任务配置包含输入、目标、生成参数、模型响应和评估指标，用于测试语言模型在复杂推理任务上的表现。

This dataset is a multi-task reasoning evaluation collection, including the ARC Challenge and multiple BBH (Big-Bench Hard) tasks such as boolean expressions, causal judgement, date understanding, disambiguation QA, formal fallacies, geometric shapes, logical deduction, movie recommendation, multistep arithmetic, navigation, object counting, table-based reasoning, and reasoning about colored objects. Each task configuration contains inputs, targets, generation arguments, model responses, and evaluation metrics, designed to assess language models performance on complex reasoning tasks.