CARTBench for Complex Multiturn Shopping Assistant Evaluation
November 3, 2026
CARTBench evaluates shopping assistants using 1,328 expert-annotated cases across 17 retail domains. It uses case-specific rubrics derived from successful real-world orders to handle complex, ambiguous requests with timelines.
HOW THIS AFFECTS YOU
●
builderYou can use these specialized rubrics to evaluate how well your shopping agents handle real-world commercial needs.