Researchers have introduced PALATE (Person-Aligned LLM-Simulated-User Assessment with Tailored Evaluation), a scalable benchmarking framework designed to address the limitations of existing role-playing agent (RPA) evaluations. Traditional methods often rely on fixed dialogue histories and static rubrics, which fail to account for how preceding context shapes RPA outputs or how individual user preferences vary. To solve this, PALATE employs a pool of 300 distinct character profiles and utilizes five specialized per-user simulators to engage candidate RPAs in free-form, multi-turn conversations. Unlike conventional benchmarks that provide a single user-independent ranking, PALATE evaluates 16 candidate models across three specific dimensions: generic turn quality, long-horizon session capability, and personalized user experience through trajectories co-constructed by the agents. A key technical contribution is the development of personalized rubrics designed to measure individual user satisfaction; empirical testing on held-out annotated data demonstrates that these tailored rubrics achieve higher agreement with human judgments than standard general rubrics. By facilitating interpretable, pair-specific evaluations of user-RPA interactions, PALATE provides a more scientifically grounded method for measuring the multi-turn conversational capabilities and emotional alignment of large language models in role-playing applications.