学术报告

学术活动

学术报告
08/15 2025 Seminar
  • Title题目 Can Theoretical Physics Research Benefit from Language Agents?
  • Speaker报告人 陆思锐/Si-Rui Lu (Max-Planck Institute of Quantum Optics)
  • Date日期 2025年8月15日 14:00
  • Venue地点 南楼6620
  • Abstract摘要

    Large Language models achieve remarkable scores through reinforcement learning on mathematics and coding. Can these systems contribute to theoretical physics beyond homework problems?Physics uniquely demands physical intuition, reasonable approximation, and consistency across extended derivations. Physicists work primarily in LaTeX, creating natural environmentsfor LLM agents—paralleling coding assistants in software development. I will share lessons learned from building TeXRA, an AI research assistant leveraging chain-of-thought reasoning, Mathematica tool integration, reflection mechanisms, and multimodal processing to maximize helpfulness and usefulness. Live demonstrations will showcase research-level tasks beyond standard benchmarks, illustrating both successes and limitations of such systems.

    Biography

    Sirui Lu is a doctoral student at Max-Planck Institute of Quantum Optics. Previously he obtained his master’s degree at the Technical University of Munich,and his bachelor from Tsinghua University. His main research interests are machine learning for quantum many-body physics.

    Inviter: Tao Shi


附件下载: