aup-anon-2026/AUP-V5-LLM-cache-Berlin-Ortsteile

AUP V5 LLM缓存（柏林Ortsteile）数据集是AUP框架V5决策引擎生成的提示-响应对集合，旨在无需消耗LLM积分即可复现NeurIPS Datasets & Benchmarks 2026提交的结果。数据集包含80,124个提示-响应对，分为基线和冲击条件，并以JSON文件形式存储，包含详细元数据。数据集为双盲评审匿名化，适用于城市规划和基于代理的建模研究。The AUP V5 LLM Cache (Berlin Ortsteile) dataset is a collection of prompt-response pairs generated by the V5 decision engine of the AUP framework, designed to reproduce results from a NeurIPS Datasets & Benchmarks 2026 submission without consuming LLM credits. It includes 80,124 prompt-response pairs, split into baseline and shock conditions, and is structured in JSON files with detailed metadata. The dataset is anonymized for double-blind review and is intended for use in urban planning and agent-based modeling research.