Molecular language model
A neural network model combining molecular structures with natural language, typically trained via contrastive learning or masked language modeling, used for multimodal tasks such as molecule-text retrieval and property prediction.
Timeline (last 90 days)
For molecular language models trained on paired structure-text data, the embedding space is not necessarily organized according to the similarity of molecular fingerprints.
Any molecular language model trained on paired structure-text data can try adding similar CKA regularization.
All Facts (2)
For molecular language models trained on paired structure-text data, the embedding space is not necessarily organized according to the similarity of molecular fingerprints.
50%UnverifiedAny molecular language model trained on paired structure-text data can try adding similar CKA regularization.
50%