Agentic programs: an emerging form of scientific software in computational materials science

📅 2026-09-01
📈 Citations: 0
Influential: 0
📄 PDF
🤖 AI Summary
本文提出了一种基于LLM的新型科学软件——代理程序,通过结合确定性算法与有限判断、任务验证及成熟过程,以解决计算材料科学中算法任务与科学判断分离的问题。
📝 Abstract
Computational materials science has traditionally delegated algorithmic tasks to computers while leaving scientific judgments to humans. We argue that recent LLM-based agent harnesses enable an emerging form of scientific software, agentic programs, that combine deterministic algorithms with bounded LLM-based judgment, task-specific verification, episodic maturation, and complete delegation in production. We illustrate this concept with DeMARS, an agentic program for constructing atomistic models from experimentally measured disordered crystal structures.
Problem

Research questions and friction points this paper is trying to address.

Computational materials science
Agentic programs
LLM-based agent
Innovation

Methods, ideas, or system contributions that make the work stand out.

agentic programs
LLM-based judgment
task-specific verification
episodic maturation
🔎 Similar Papers
No similar papers found.
Y
Yunsung Lim
Research Institute of Advanced Materials, Seoul National University, Seoul, 08826, Republic of Korea
H
Haekwan Jeon
Department of Materials Science and Engineering, Seoul National University, Seoul, 08826, Republic of Korea
J
Jaesun Kim
Department of Materials Science and Engineering, Seoul National University, Seoul, 08826, Republic of Korea
Jisu Kim
Jisu Kim
Seoul National University
S
Seungwu Han
Research Institute of Advanced Materials, Seoul National University, Seoul, 08826, Republic of Korea; Department of Materials Science and Engineering, Seoul National University, Seoul, 08826, Republic of Korea; Center for AI and Natural Sciences, Korea Institute for Advanced Study, Seoul, 02455, Republic of Korea