| 11 | 0 | 187 |
| 下载次数 | 被引频次 | 阅读次数 |
高能物理实验研究周期长、知识门槛高，受限于人力，大量对撞数据难以得到充分的挖掘与分析。以大语言模型为基础的智能体能够编写代码、检索文献并执行多步分析，为突破困境提供契机。本文聚焦Just Furnish Context(JFC)智能体框架，分析其完成的CMS实验H→τ+τ~-信号强度测量、ALEPH实验Lund喷注平面密度测量等结果，及其对高能物理研究的意义；同时概括其当前局限性，并展望基准数据集构建、下一代大科学装置数据处理、人才培养的发展前景。
Abstract：Particle physics experiments have long research cycles and high barriers to entry.Limited human resources make it difficult to fully explore and analyze large volumes of collision data.Agents based on large language models can write code,retrieve literature,and carry out multi-step analyses,offering an opportunity to address this challenge.This article focuses on the Just Furnish Context(JFC)agent framework and examines results produced by the framework,including the CMS H→τ+τ~-signal-strength measurement and the ALEPH Lund jet-plane density measurement,as well as their implications for particle physics research.It also summarizes the framework's current limitations and discusses prospects for benchmarkdataset construction,data processing at next-generation large-scale scientific facilities,and the training of future physicists.
[1] GUEST D,CRANMER K,WHITESON D.Deep learning and its application to LHC physics[J].Annual Review of Nuclear and Particle Science,2018,68:161-181.
[2] GENDREAU-DISTLER E,HO J,KIM D,et al.Automating high energy physics data analysis with LLM-powered agents[Z].arXiv:2512.07785,2025.
[3] BADEA A,CHEN Y,MAGGI M,et al.Agentic AI-physicist collaboration in experimental particle physics:a proof-of-concept measurement with LEP open data[Z].arXiv:2603.05735,2026.
[4] LI K,LIU B J,MELLADO B,et al.AI agents,language,deep learning,and the next revolution in science[J].Frontiers of Physics,2026,21(9):096401.
[5] MORENO E A,BRIGHT-THONNEY S,NOVAK A,et al.AI agents can already autonomously perform experimental high energy physics[Z].arXiv:2603.20179,2026.
[6] MCGREIVY J,DELANEY B,BECK A,et al.Seeing the forest through the trees:knowledge retrieval for streamlining particle physics analysis[Z].arXiv:2509.06855,2025.
[7] The CMS Collaboration.Evidence for the 125GeV Higgs boson decaying to a pair of τ leptons[J].Journal of High Energy Physics,2014,2014(5):104.
[8] ATLAS Collaboration.Measurement of the Lund jet plane using charged particles in 13TeV proton-proton collisions with the ATLAS detector[J].Physical Review Letters,2020,124:222002.
[9] The ALEPH Collaboration.Measurement of the Z resonance parameters at LEP[J].The European Physical Journal C-Particles and Fields,2000,14(1):1-50.
[10] HENDRYCKS D,BURNS C,KADAVATH S,et al.Measuring mathematical problem solving with the MATH dataset[C]//Proceedings of the 35th Conference on Neural Information Processing Systems Track on Datasets and Benchmarks.2021.
[11] JIMENEZ C E,YANG J,WETTIG A,et al.The Twelfth International Conference on Learning Representations,2024.
[12] FAROUGHY D A,SCHWEITZER S P,PANG I,et al.Collider-Bench:benchmarking AI agents with particle physics analysis reproduction[Z].arXiv:2605.13950,2026.
[13] DIERLAMM A,on behalf of the CMS Tracker Collaboration.The CMS outer tracker upgrade for the HL-LHC[J].Nuclear Instruments and Methods in Physics Research Section A:Accelerators,Spectrometers,Detectors and Associated Equipment,2019,924:256-261.
[14] BOCCALI T.Computing models in high energy physics[J].Reviews in Physics,2019,4:100034.
[15] FCC Collaboration.FCC-ee:The lepton collider[J].The European Physical Journal Special Topics,2019,228(2):261-623.
[16] The CEPC Study Group.CEPC technical design report:Accelerator[J].Radiation Detection Technology and Methods,2024,8(1):1-1105.
[17] AI X C,AN L P,AN S Z,et al.Conceptual design report of the Super Tau-Charm Facility:The accelerator[J].Nuclear Science and Techniques,2025,36(12):242.
基本信息:
DOI：10.27024/j.wlygc.2026.08.13.is
中图分类号:TP18;O572
引用信息:
[1]赵一扬,胡震.以人工智能窥探粒子的奥秘[J].物理与工程,2026,46(05):38-44.DOI:10.27024/j.wlygc.2026.08.13.is.
基金信息:
国家自然科学基金项目W2511007、125B1008; 清华大学自主科研计划的支持
2026-08-25
2026-08-25