AutoLogi
收藏资源简介:
AutoLogi是一个自动生成的开放性问题逻辑谜题数据集,由复旦大学计算机科学技术学院和阿里巴巴集团共同创建。该数据集通过程序化验证和可控难度等级,旨在为大型语言模型提供更可靠的推理能力评估。数据集包含1575个英文逻辑谜题和883个中文逻辑谜题,全部由高级语言模型生成,并经过验证函数检查以确保正确性。该数据集的应用领域是逻辑推理能力的评估,旨在解决现有标准多项选择题格式容易导致随机猜测的问题。
AutoLogi is an automatically generated open-ended logical puzzle dataset co-created by the School of Computer Science, Fudan University and Alibaba Group. Equipped with programmatic verification and controllable difficulty levels, this dataset is designed to provide more robust evaluations of reasoning capabilities for large language models. It comprises 1,575 English logical puzzles and 883 Chinese logical puzzles, all generated by advanced large language models and validated via dedicated validation functions to ensure their correctness. Focused on logical reasoning ability evaluation, this dataset aims to resolve the problem that existing standard multiple-choice question formats are highly prone to random guessing.

- 1AutoLogi: Automated Generation of Logic Puzzles for Evaluating Reasoning Abilities of Large Language Models复旦大学计算机科学技术学院, 阿里巴巴集团 · 2025年



