Sparse Feature Masks Enhance Molecular Toxicity Prediction with Chemical Language Models

This research explores a novel application of sparse feature masks within chemical language models for predicting molecular toxicity, a critical area in drug discovery and environmental science. The use of sparse masks likely improves model interpretability and efficiency by focusing on the most relevant chemical features.

• •Applies sparse feature masks to chemical language models.
• •Focuses on predicting molecular toxicity.
• •Potentially improves model efficiency and interpretability.