Chemical Language Models

This section covers the application of language model architectures to chemistry. Most notes here focus on models that treat molecular strings (SMILES, IUPAC names, InChI) as sequences, including encoder-only transformers like ChemBERTa and sequence-to-sequence models like ChemFormer and STOUT for structure-to-name translation. A growing set of notes also covers multimodal approaches: models like ChemVLM and InstructMol that connect 2D molecular images or graphs with natural language, enabling tasks like captioning, question answering, and structure retrieval.

Computational Chemistry

Diagram showing Struct2IUPAC workflow: molecular structure (SMILES) passing through Transformer to generate IUPAC name, with round-trip verification loop

Struct2IUPAC: Translating SMILES to IUPAC via Transformers

This paper proposes a Transformer-based approach (Struct2IUPAC) to convert chemical structures to IUPAC names, challenging the dominance of rule-based systems. Trained on ~47M PubChem examples, it achieves near-perfect accuracy using a round-trip verification step with OPSIN.

Computational Chemistry

Transformer encoder-decoder architecture processing InChI string character-by-character to produce IUPAC chemical name

Translating InChI to IUPAC Names with Transformers

This study presents a sequence-to-sequence Transformer model that translates InChI identifiers into IUPAC names character-by-character. Trained on 10 million PubChem pairs, it achieves 91% accuracy on organic compounds, performing comparably to commercial software.

Computational Chemistry

ChemVLM architecture showing molecular structure and text inputs flowing through vision encoder and language model into multimodal LLM for chemical reasoning

ChemVLM: A Multimodal Large Language Model for Chemistry

A 2025 AAAI paper introducing ChemVLM, a domain-specific multimodal LLM (26B parameters). It achieves state-of-the-art performance on chemical OCR, reasoning benchmarks, and molecular understanding tasks by combining vision and language models trained on curated chemistry data.