●builderYou should look beyond semantic similarity when building personalization engines to avoid failure in non-obvious preference scenarios.
●researcherThe benchmark identifies a gap in how models handle profile-preference conceptual misalignment (PRCM).