Serin-kimm/Persona2Web

Persona2Web是一个用于评估个性化网络代理在真实开放网络上性能的基准数据集。它包含多个数据组件：ground_truth.jsonl提供所有任务的真实个性化信息；query_personalization_0.jsonl、query_personalization_1.jsonl和query_personalization_2.jsonl分别包含不同模糊级别的查询任务，从明确约束到完全隐式约束；user_history.jsonl包含用户历史数据，通过行为模式隐式揭示偏好。数据集还包括子目录中的JSON文件，用于具体任务评估和历史记录。整体上，该数据集旨在通过用户历史进行上下文推理，测试代理在个性化查询中的表现。Persona2Web is a benchmark for evaluating personalized web agents on the real open web. It includes multiple data components: ground_truth.jsonl provides ground-truth personalization information for all tasks; query_personalization_0.jsonl, query_personalization_1.jsonl, and query_personalization_2.jsonl contain queries at different ambiguity levels, ranging from explicit constraints to fully masked constraints; user_history.jsonl contains user histories that reveal preferences implicitly through behavioral patterns. The dataset also includes JSON files in subdirectories for task evaluation and historical records. Overall, it aims to test agents performance in personalized queries through contextual reasoning with user history.