Target context boosts performance in scarce-data regimes when fused properly via FiLM but degrades results under distribution shift, while standard molecular benchmarks suffer from severe leakage and trivial baselines.
Ethan Perez, Florian Strub, Harm de Vries, Vincent Dumoulin, and Aaron Courville
5 Pith papers cite this work. Polarity classification is still indexing.
5
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 5roles
background 1polarities
background 1representative citing papers
Molexar is a unified multimodal molecular foundation model built on Fragment-SELFIES that uses pretraining followed by supervised fine-tuning with in-place condition embedding to handle scalar properties, pharmacophores, proteins, and pockets in one autoregressive path.
An audit of 26 papers using the MassSpecGym benchmark finds evaluation failures in at least 17, including data leakage and metric divergence, and releases an updated v1.5 suite with fixes.
Boltz-2 and fine-tuned DrugFormDTA lead ML-based binding prediction while GNINA leads docking tools on a cleaned antiviral dataset, with performance varying by viral protein.
On LIT-PCBA, supervised ML re-ranking reaches median EF1% of 4.49 while AutoDock-GNINA scores 2.14 and consensus does not exceed the best single scorer.
citing papers explorer
-
When Does Context Help? A Systematic Study of Target-Conditional Molecular Property Prediction
Target context boosts performance in scarce-data regimes when fused properly via FiLM but degrades results under distribution shift, while standard molecular benchmarks suffer from severe leakage and trivial baselines.
-
Molexar: A Unified Multimodal Molecular Foundation Model for Drug Design
Molexar is a unified multimodal molecular foundation model built on Fragment-SELFIES that uses pretraining followed by supervised fine-tuning with in-place condition embedding to handle scalar properties, pharmacophores, proteins, and pockets in one autoregressive path.
-
MassSpecGym in the Wild: Uncovering and Correcting Evaluation Pitfalls in AI-Driven Molecule Discovery
An audit of 26 papers using the MassSpecGym benchmark finds evaluation failures in at least 17, including data leakage and metric divergence, and releases an updated v1.5 suite with fixes.
-
Benchmarking open-source tools for in silico antiviral drug discovery
Boltz-2 and fine-tuned DrugFormDTA lead ML-based binding prediction while GNINA leads docking tools on a cleaned antiviral dataset, with performance varying by viral protein.
-
Benchmarking Single-Pose Docking, Consensus Rescoring, and Supervised ML on the LIT-PCBA Library: A Critical Evaluation of DiffDock, AutoDock-GPU, GNINA, and DiffDock-NMDN
On LIT-PCBA, supervised ML re-ranking reaches median EF1% of 4.49 while AutoDock-GNINA scores 2.14 and consensus does not exceed the best single scorer.