Tag
The Shopping Reasoning Bench is an expert-authored benchmark for evaluating multi-turn conversational shopping assistants, with 525 missions and over 10,000 binary rubrics. Evaluations of GPT, Claude, and Gemini show that current models achieve only 57-77% pass rates, revealing significant gaps in expert-level shopping reasoning.