research
The Molecule Builder Added New Fragments Without Retraining
Fraglingo jointly predicts fragment identity and attachment and generalized to inference libraries four times larger than its training vocabulary.
Summary
Fraglingo jointly predicts fragment identity and attachment and generalized to inference libraries four times larger than its training vocabulary.
Fragment-based molecular generators often choose a fragment from a fixed vocabulary and predict its attachment separately. Fraglingo represents both decisions in one attachment-aware continuous embedding, retrieves the next fragment by nearest-neighbor search and encodes the growing molecule from its active attachment site. Because inference operates over embeddings rather than fixed identifiers, new fragments can be added without retraining. The authors report stronger joint property control than comparable baselines and generalization to fragment libraries four times larger than the training set. Computational benchmarks do not establish laboratory synthesis or drug efficacy.
Why it matters
Fraglingo jointly predicts fragment identity and attachment and generalized to inference libraries four times larger than its training vocabulary.
Limits and context
- Computational benchmarks do not establish laboratory synthesis or drug efficacy.
Key claims
Fraglingo jointly predicts fragment identity and attachment and generalized to inference libraries four times larger than its training vocabulary.
Qualification: Computational benchmarks do not establish laboratory synthesis or drug efficacy.
Evidence: source-2026-09-15-010
Sources
- arXiv preprint 2609.13519arXiv · primary research
Corrections
No corrections have been recorded for this story.