BAZIQA-BENCHMARK
收藏资源简介:
BAZIQA-BENCHMARK是由上海交通大学与上海指南信息技术联合构建的专业级八字命理推理评测基准,包含200道来自2021-2025年全球算命师竞赛的多项选择题。数据集涵盖婚姻、职业、健康等七大应用领域,每道题目均基于预计算的标准化命盘结构,要求模型进行符号推理与时间条件组合分析。数据通过竞赛组委会专业筛选并匿名化处理,旨在评估大语言模型在非标准符号系统下的结构化推理能力,为玄学与传统文化领域的AI研究提供可复现的量化基准。
BAZIQA-BENCHMARK is a professional Bazi numerology reasoning evaluation benchmark co-developed by Shanghai Jiao Tong University and Shanghai Zhinan Information Technology. It contains 200 multiple-choice questions from the 2021-2025 Global Fortune-Telling Competition. The dataset covers seven application domains including marriage, career, health and others. Each question is based on a pre-calculated standardized natal chart structure, requiring the model to conduct symbolic reasoning and temporal conditional combinatorial analysis. The data has been professionally screened and anonymized by the competition organizing committee. This benchmark aims to evaluate the structured reasoning capability of large language models (LLMs) under non-standard symbolic systems, and provide a reproducible quantitative benchmark for AI research in the fields of metaphysics and traditional culture.



