This research introduces a multimodal large language model that can automatically parse chemical reaction images into structured data, reducing the need for manual curation of chemical information.

• Addresses a critical bottleneck in chemical knowledge processing
• Enables the conversion of published reactions into machine-readable formats
• Provides a scalable approach to build comprehensive chemical databases
• Supports broader application of AI in organic chemistry research

For Biology and Medical fields, this technology allows researchers to rapidly access and analyze vast chemical reaction data from literature, accelerating drug discovery and biochemical research processes.

Towards Large-scale Chemical Reaction Image Parsing via a Multimodal Large Language Model