38‐1: Enhancing Chemical Capabilities of Large Language Models for OLED Materials Design
SID Symposium Digest of Technical Papers · 2026-08 · 57 (S1) · 348-351
DOI: 10.1002/sdtp.70003_91 ↗Kurz gefasst
Fachartikel in der Zeitschrift SID Symposium Digest of Technical Papers, online erschienen am 11.09.2026. Verfasst von Tsun-Hin Cheung und 4 weiteren Autor:innen. Die Arbeit nennt 20 Literaturangaben.
Abstract (Originalsprache)
We introduce a domain‐specific Large Language Model (LLM) post‐trained to enable interactive design of Organic Light‐Emitting Diode (OLED) materials through molecular generation, property optimization, and structural editing using natural language. Our strategy includes curating a large dataset of text‐represented OLED‐relevant molecules with accurate HOMO/LUMO energies and constructing diverse instruction‐following pairs for Supervised Fine‐Tuning (SFT), along with thousands of synthetic questions for Reinforcement Learning (RL) with chemical feedback. We incorporate three auxiliary models—a validity verifier, a novelty classifier, and a property predictor—to collectively improve the validity, novelty, and accuracy of the designed molecules. The trained LLM processes natural‐language instructions from chemists to generate multiple valid, novel, and property‐optimized molecular candidates, significantly streamlining the design‐make‐test cycle for next‐generation OLED materials. Notably, RL‐based post‐training substantially improves performance over the SFT‐only baseline. This framework demonstrates strong potential in OLED materials design and paves the way for AI‐driven development of high‐performance organic semiconductors.
Zitieren
APA
Cheung, T., Xu, W., Lin, X., Wu, Z., & Ma, L. (2026). 38‐1: Enhancing Chemical Capabilities of Large Language Models for OLED Materials Design. SID Symposium Digest of Technical Papers, 57(S1), 348-351. https://doi.org/10.1002/sdtp.70003_91
Vancouver
Cheung T, Xu W, Lin X, Wu Z, Ma L. 38‐1: Enhancing Chemical Capabilities of Large Language Models for OLED Materials Design. SID Symposium Digest of Technical Papers. 2026;57(S1):348-351. doi:10.1002/sdtp.70003_91
BibTeX
@article{Cheung2026,
author = {Cheung, Tsun-Hin and Xu, Wei and Lin, Xinshi and Wu, Zekai and Ma, Lan},
title = {38‐1: Enhancing Chemical Capabilities of Large Language Models for OLED Materials Design},
journal = {SID Symposium Digest of Technical Papers},
year = {2026},
volume = {57},
number = {S1},
pages = {348--351},
doi = {10.1002/sdtp.70003_91}
}
EUVIMED