소재 AI 공유함
Menu
← AI in materials

Materials AI · written by AI

A method for dynamically defining the search space with an LLM in autonomous materials discovery

AI editorPublished 2026.08.25

A Communications Materials paper addressed the bottleneck of defining the search space in autonomous closed-loop materials discovery by dynamically adjusting it with an LLM. It can be seen as an example of AI managing the search range at the intersection of autonomous experimental workflows and materials informatics. [1]

1/ AI use method: hierarchical search combining a large language model with Bayesian optimization.
2/ Applied task: automatically resetting the candidate space during materials search and selecting the next candidate.
3/ Reported result: the improvement in search performance was most pronounced for target properties for which the LLM had abundant prior knowledge.
4/ The outer loop LLM updates the materials space to be searched, and the inner loop Bayesian optimization selects candidates within that space.
5/ This approach reduces the burden of repeatedly specifying the search region by hand.
6/ Performance depends on the target property and the model’s relevant knowledge.
7/ The result should not be expanded into a claim of equal improvement across all material families or complete experimental autonomy. [1]

How to apply this

You can try a small-scale closed-loop materials search workflow that dynamically updates the search space with an LLM.

  1. Define the initial materials data and target property, and also set the upper and lower bounds of the search range that a person will allow.
  2. Create a rule by which the LLM updates the materials space to be examined next based on the current data and target.
  3. Within the updated space, use Bayesian optimization to choose the next candidate and verify performance with one or two experiments or calculations.
  4. Feed the new result back in to redefine the search space, and compare candidate diversity and hit rate with the baseline workflow.
Data needed
Candidate composition or structure lists, existing experimental or computational results, the target property definition, and initial materials data for the search are needed.
Checking the result
Check whether the search space updated by the LLM and the candidates recommended by Bayesian optimization present more diverse promising candidates than the baseline workflow, and whether the predictions hold in follow-up experiments or calculations.
What can go wrong
The most common problem is that the LLM’s prior knowledge may wrongly narrow the search space for weakly known properties or unfamiliar material systems, inflating apparent performance.

이 가이드는 원문의 방법을 바탕으로 AI가 제안한 적용 방향입니다. 원문에 적힌 내용이 아니며 같은 결과가 재현된다는 보장도 아닙니다.

Related concepts and practice

이 이슈를 읽는 데 필요한 개념입니다. 용어마다 Colab에서 바로 돌려볼 수 있는 실습 노트북이 있습니다.

Sources

  1. Large-language-model-driven adaptive search space definition for autonomous closed-loop materials exploration | Communications Materials

웹 자료를 바탕으로 AI가 자동 요약한 글입니다. 연구의 세부 조건과 결과는 연결된 원문을 확인해주세요.