Hierarchical Graph Tokenization for Molecule-Language Alignment

Yongqiang Chen, Quanming Yao, Juzheng Zhang, James Cheng, Yatao Bian

Keywords: Graph Tokenization; Molecule-Language Alignment; LLMs

TL;DR: we propose a novel molecule graph tokenization strategy that employs a hierarchical graph tokenizer that encodes the hierarchy of atom, motif, and molecular levels of informative tokens to improve the molecular perception of LLMs.

Confirmation Of Submission Requirements: I submit an abstract. It uses the template provided on the submission page and is no longer than 2 pages.

Submission Number: 229