Ruoyu Zhang, Zhaocen Liu, Weiyi Tian, Melissa Feng, Shilong Liu · 2026 · arXiv
DOI: 10.48550/arXiv.2610.28491 · open-access PDF
Los asistentes a largo plazo deben seguir tareas inacabadas a lo largo de los días a medida que cambien la información y las circunstancias. Encontrar información o comenzar una acción, sin embargo, no garantiza que la tarea se complete. Introducimos OptionWorld, un índice que rastrea la adquisición de información, juicios, acciones y resultados en tareas simuladas de la vida diaria para identificar dónde los LLM actuales caen cortos como asistentes personales a largo plazo. En las entradas de decisión archivadas, los hechos clave permanecen en la última entrada del modelo antes del plazo en el 35,5% de los casos fallidos.
Machine translation (Meta M2M100); the English original follows.
Long-term assistants must follow unfinished tasks across days as information and circumstances change. Finding information or starting an action, however, does not ensure that the task gets completed. We introduce OptionWorld, a benchmark that traces information acquisition, judgments, actions, and outcomes in simulated daily-life tasks to identify where current LLMs fall short as long-term personal assistants. In runs with archived decision inputs, the key facts remain in the model's last input before the deadline in 35.5% of failed instances. Failure analysis identifies the three most freque
Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese
GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.28491, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txtSources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange