一句话：提出分层多智能体 LLM 系统 ScholForge，端到端自主完成软件工程研究并产出代码、实验与论文。
软件工程研究涵盖文献调研、软件构建、实验运行与论文写作等多个阶段，通常各阶段需要不同工具和专业知识，难以端到端自动化。本文提出 ScholForge，一个分层多智能体 LLM 系统：Manager 智能体通过 Code-Act 执行循环将任务委派给 Librarian、Proposer、MLE Solver、Writer 和 Reviewer 五个专家智能体，并通过三个 Model Context Protocol（MCP）服务器暴露外部能力。给定自然语言研究目标，系统无需人工干预即可生成可运行代码、实验结果与编译好的论文手稿。作者展示了一个代表性案例 Prompt2Edge：该系统自主开发了面向边缘设备的 agentic ML 流水线，在 Speech Commands 上以 17.8k 参数模型取得 86.71% 测试准确率，在 UCI HAR 上取得 0.9638 Macro-F1，消耗约 5M tokens，并在 5 小时内完成整个流程。
Software engineering research spans many activities: surveying literature, building software, running experiments, writing papers-that typically require different tools and expertise at each stage. We present a ScholForge - hierarchical multi-agent LLM system that tackles this end-to-end. Manager agent delegates to five specialists (Librarian, Proposer, MLE Solver, Writer, Reviewer) through a Code-Act execution loop, with external capabilities exposed via three Model Context Protocol (MCP) servers. Given a natural-language research goal, the system produces working code, experimental results, and a compiled manuscript without human intervention. We present one representative case study: Prompt2Edge, an autonomously developed agentic ML pipeline for edge devices that achieved 86.71% test accuracy on Speech Commands with a 17.8k-parameter model and 0.9638 Macro-F1 on UCI HAR, consuming approximately 5 M tokens and completing the full pipeline in under 5 hours.
@inproceedings{StepinTKKKMGM26,
title = {ScholForge: A Multi-Agent LLM System for Autonomous Software Engineering Research},
author = {Aleksei Stepin and Boris Tolstokulakov and Vladislav Kulikov and Artyom Kabanov and Vladimir Kubasov and Mikhail Mozikov and Ilya Gusev and Ilya Makarov},
booktitle = {FSE},
year = {2026},
doi = {10.1145/3803437.3807394}
}