LATIDIA · Investigación

Un rasgo de comportamiento se filtra en las preferencias: diagnóstico de la interferencia de rasgos en los simuladores de usuarios de LLM

arXiv: 2609.25572v1Announce Type: new Resumen: Los simuladores de usuario basados en LLM tienen como objetivo cerrar la brecha fuera de línea en la evaluación de recomendadores emulando a los usuarios a través de rasgos inyectados, donde los atributos de preferencia determinan w

La noticia

arXiv:2609.25572v1 Announce Type: new Abstract: LLM-based user simulators aim to bridge the offline-online gap in recommender evaluation by emulating users through injected traits, where preference attributes determine what a user engages with and a behavioral activity trait governs how long they browse. However, we show this intended trait independence collapses during simulation, causing two failures: (i) Trait Interference, where amplified activity distorts preference boundaries and forces interactions with mismatched items to sustain browsing, and (ii) Evaluation Invalidity, where satisfaction scores inflate with activity-driven page counts despite taste mismatches, biasing evaluation toward trait distributions rather than recommender performance. To resolve this, we propose PQA, a page-level quality anchoring method that guides