通过层级集体推理扩展用于材料设计的LLM智能体
Scaling LLM Agents for Materials Design through Hierarchical Collective Reasoning
- Seoul National University（首尔大学）
机构由 AI 辅助整理，请以论文原文为准。
中文总结 AI 辅助
HiMatGen通过层级集体推理框架扩展LLM智能体，利用多领域讨论小组和工具代表协作，显著提升材料设计中的晶体候选生成效率与属性满足能力。
中文摘要 AI 辅助
材料设计必须协调相互竞争的功能需求与稳定性和合成约束。生成模型可以产生稳定且新颖的晶体，但容纳详细的自然语言设计指令仍然具有挑战性。在此，我们引入HiMatGen，一个通过层级集体推理来扩展大型语言模型智能体的框架。HiMatGen基于GPT-5.6 Terra构建，将十个科学领域的讨论小组中的100名调查员与支持工具领域的代表连接起来。代表们调查提案，跨领域交换证据，并将发现和未解决的问题返回给各自的小组。这种双向交流将专家的分歧转化为结构修订和替代设计。在六个化学体系中，HiMatGen产生的最终晶体候选数量是支持工具的GPT-6 Astra单智能体基线的1.9倍，产生的稳定、独特且新颖（SUN）结构的数量是其1.7倍。在属性导向的任务中，完整的HiMatGen工作流优于同模型单智能体和独立生成对照。与MatterGen和Chemeleon2的比较表明，它能够开发满足联合功能和化学要求的多样化晶体。在从相同提案初始化的配对比较中，层级讨论将辩论阶段的令牌使用量相对于完整辩论减少了22倍。HiMatGen提供了一种材料设计方法，随着智能体群体的增长，在迭代探索过程中维持科学知识和计算证据的交流。
英文摘要
Materials design must reconcile competing functional requirements with stability and synthesis constraints. Generative models can produce stable, novel crystals, but accommodating detailed natural-language design instructions remains challenging. Here we introduce HiMatGen, a framework for scaling large language model agents through hierarchical collective reasoning. Built from GPT-5.6 Terra, HiMatGen connects 100 investigators in discussion pods across ten scientific domains with tool-enabled domain representatives. Representatives investigate proposals, exchange evidence across domains and return findings and unresolved questions to their pods. This bidirectional exchange turns specialist disagreements into structural revisions and alternative designs. Across six chemical systems, HiMatGen produces 1.9 times as many final crystal candidates and 1.7 times as many stable, unique and novel (SUN) structures as a tool-enabled GPT-6 Astra single-agent baseline. In property-directed tasks, the complete HiMatGen workflow outperforms same-model single-agent and independent-generation controls. Comparisons with MatterGen and Chemeleon2 demonstrate its ability to develop diverse crystals satisfying joint functional and chemical requirements. In a paired comparison initialized from the same proposals, hierarchical discussion reduces debate-stage token usage by a factor of 22 relative to full debate. HiMatGen provides an approach to materials design that sustains the exchange of scientific knowledge and computational evidence throughout iterative exploration as the agent population grows.
