[15p-S2_204-1~18] 23.1 Joint Session N "Informatics"

4:00 PM - 4:15 PM JST(7:00 AM - 7:15 AM UTC)Presentation by Applicant for JSAP Young Scientists Presentation Award

[15p-S2_204-10] LLM-Based Symbolic Search for Interpretable Formulas of Experts’ Knowledge: Application to SiC Solution Growth

〇(M1)Kota Takahara¹, Kentaro Kutsukake^1,2, Shunta Harada^1,2, Toru Ujihara^1,2 (1.Nagoya Univ., 2.IMaSS Nagoya Univ.)

Keywords:

LLM,Implicit Knowledge

In SiC solution growth optimization, formulating metrics based on experts' tacit knowledge is challenging, and pairwise comparison scores often lack clear rationale. We investigated a method using LLMs to explore and formulate the determinants of these scores. Validation showed that for reasoning models, appropriate instructions stabilized the exploration, leading to the successful identification of correct judgment axes. This suggests the method is effective for explicitizing implicit evaluations into interpretable formulations.