2026-10-14T07:43:03Z https://www.scidb.cn/oai

10.57760/sciencedb.j00214.00209 2026-05-14T10:45:29Z

2026-05-14 Evaluation of Reasoning and Thinking Abilities of Large Language Models in Economic Problems (Appendix) doi:10.57760/sciencedb.j00214.00209 en This dataset is the appendix of the identified paper. Experiment 1.1 consists of the Intuitive Alignment group and the Intuitive Contradiction group (200 questions each), testing whether the models rely on real-world knowledge bias. Experiment 1.2 is a control experiment with varied narrative order (200 questions each), examining the models’ sensitivity to text arrangement order. The tested models cover mainstream series such as Llama, Gemma, and Mistral. large language models; perturbation testing; evaluation of reasoning robustness wang yong yi Cheng Bing PUBLIC https://creativecommons.org/licenses/by/4.0/ dataset Science Data Bank