A scaffold-prompted GPT with two-phase pretraining, reinforcement fine-tuning, and reward-guided decoding reports improved drug-optimization scores on COVID and cancer benchmarks, though evaluation and training share the same molecules and a surrogate docking score.
Mol GPT : Molecular generation using a transformer-decoder model
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
q-bio.BM 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
ScaffoldGPT: A Scaffold-based GPT Model for Drug Optimization
A scaffold-prompted GPT with two-phase pretraining, reinforcement fine-tuning, and reward-guided decoding reports improved drug-optimization scores on COVID and cancer benchmarks, though evaluation and training share the same molecules and a surrogate docking score.