yuzhan2205/Shopping-companion

Shopping Companion Shopping Companion is a benchmark and training resource for long-horizon, preference-grounded e-commerce agents. It evaluates whether a tool-using agent can recover a user's preferences from cross-session conversation history and apply those preferences while searching and inspecting a large real-world product catalog. The benchmark contains two task types: Single-product recommendation: retrieve the relevant long-term preference and find one product that… See the full description on the dataset page: https://huggingface.co/datasets/yuzhan2205/Shopping-companion.