LATIDIA · Ciberseguridad
Orbit AI: Evaluación de agentes potenciados por herramientas para responder a preguntas de empresas
arXiv:2611.19333v1 Tipo de anuncio: nuevo Resumen: La información empresarial suele almacenarse en varias plataformas en lugar de en una única ubicación centralizada. Responder a preguntas sobre las actividades en el lugar de trabajo puede requerir información
WhatsApp ↗Telegram ↗
La noticia
arXiv:2611.19333v1 Announce Type: new Abstract: Enterprise information is commonly stored across multiple platforms rather than in one centralized location. Answering questions about workplace activities may require information from several sources, including project documents, task management systems, and team conversations. Therefore, we evaluate Orbit AI, a tool-based retrieval-augmented generation (RAG) system that uses a ReAct agent implemented with LangGraph to retrieve and combine information from three different workspace sources. We compare three configurations: (1) a standalone LLM without retrieval, (2) document-based RAG, and (3) a multi-tool agent with access to all three sources. The held-out evaluation set contains 80 questions designed for Phase 3 workspace; Gemini 3.5 Flash Lite achieves an