כתבה
arXiv cs.AI ·
צורות של אינטגרציה של LLM באפליקציות
Forms of LLM-Integrated Applications from LLM-Chats to Autonomous AI Agent System
מאמר זה מסקר צורות של אינטגרציה של LLM באפליקציות, כולל צורות שונות של כלים ומודלים כמו Claude, Gemini ו-LangGraph.
תקציר מקורי באנגליתarXiv:2610.11899v1 Announce Type: cross Abstract: Large language models (LLMs) are increasingly embedded as components in software systems, marketed under labels such as chatbot, copilot, retrieval-augmented generation, workflow, coding agent and AI agent. Whether these labels denote genuine architectural forms or serve as branding has not been assessed systematically. In the sources surveyed, labels do carry architectural content, most clearly in vendor usage: copilot denotes a router-worker architecture operating a host application under step-by-step user confirmation, while the more recent shift to the label agent coincides with AI-planned multi-step execution of which the user sees only the outcome. The coding agents of four major providers share one architecture, a reason-and-act loop
קרא במקור המקורי
arxiv.org
פתח כתבה מקורית