CatalyticMLLM: A Graph-Text Multimodal Large Language Model for Catalytic Materials
The paper introduces QE-Catalytic-V2, a multimodal large language model designed for catalytic materials. This model integrates property prediction and inverse design into a unified framework, enhancing the efficiency of the optimization process. Experimental results indicate that QE-Catalytic-V2 outperforms traditional decoupled approaches in both prediction and design tasks.
- ▪QE-Catalytic-V2 is a unified graph-text multimodal large language model for catalytic materials.
- ▪The model integrates property prediction and inverse design within the same representation space.
- ▪Experimental results show that this unified approach outperforms decoupled baselines in catalytic tasks.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Artificial Intelligence arXiv:2605.17254 (cs) [Submitted on 17 May 2026] Title:CatalyticMLLM: A Graph-Text Multimodal Large Language Model for Catalytic Materials Authors:Yanjie Li View a PDF of the paper titled CatalyticMLLM: A Graph-Text Multimodal Large Language Model for Catalytic Materials, by Yanjie Li View PDF HTML (experimental) Abstract:Property prediction and inverse structural design of catalytic materials are typically modeled as two independent tasks: the former predicts target properties from given structures, whereas the latter generates candidate structures according to desired properties.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
