FARM model enhances molecular representations with functional group awareness

Researchers have developed a new foundation model called FARM (Functional Group-Aware Representations for Small Molecules) to improve how AI understands molecular structures. FARM incorporates functional group annotations at the atomic level, creating enhanced SMILES strings and molecular graphs that capture both chemical knowledge and structural connectivity. This approach allows Transformer-based models to better process molecular data, leading to state-of-the-art performance on 8 out of 13 tasks in the MoleculeNet benchmark. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Enhances molecular representation learning, supporting transfer learning in drug discovery and materials science.

RANK_REASON This is a research paper introducing a new model and benchmark results.

Read on arXiv cs.LG →

• FARM
• MoleculeNet
• Transformer

• paper
• model release