论文概要
研究领域: CV 作者: Yunbo Long 发布时间: 2026-03-25 arXiv: 2603.24402
中文摘要
现有的自动化研究系统作为无状态的线性流水线运行，生成输出时无法持续理解研究图景。它们顺序处理论文、提出想法而缺乏结构化的差距分析，并且缺乏智能体验证或改进彼此发现的机制。本文提出AutoProf（自主教授），一种多智能体编排框架，其中专业智能体提供端到端的AI研究监督，由人类兴趣驱动，从文献综述到差距发现、方法开发、评估和论文写作。
原文摘要
Existing automated research systems operate as stateless, linear pipelines, generating outputs without maintaining a persistent understanding of the research landscape. They process papers sequentially, propose ideas without structured gap analysis, and lack mechanisms for agents to verify or refine each other's findings. We present AutoProf (Autonomous Professor), a multi-agent orchestration framework where specialized agents provide end-to-end AI research supervision driven by human interests, from literature review through gap discovery, method development, evaluation, and paper writing.
--- *自动采集于 2026-03-27*
#论文 #arXiv #CV #小凯