analiza aiqbench
MolSight demonstrates that vision encoders can effectively extract chemical insights directly from 2D skeletal diagrams of molecules, offering a computationally efficient alternative to heavy graph or language models. By utilizing a chemistry-informed curriculum on two million images, this approach achieves top-tier performance across physical-property regression and drug-discovery tasks while reducing computational costs by an order of magnitude compared to multi-modal competitors. This work highlights the viability of computer vision techniques for solving complex problems in molecular science without relying on billion-parameter architectures.
MolSight: Molecular Property Prediction with Images
arXiv:2605.10157v2 Announce Type: replace-cross Abstract: Every molecule ever synthesised can be drawn as a 2D skeletal diagram, yet in modern property prediction this universally available representation has received less focus in favour of molecular graphs, 3D conformers, or billion-parameter language models, each imposing its own computational and data-engineering overhead. We present $\textbf{MolSight}$, the first systematic large-scale study of vision-based Molecular Property Prediction (MPP). Using 10 vision architectures, 7 pre-training strategies, and $2\,M$ molecule images, we evaluate performance across 10 downstream tasks spanning physical-property regression, drug-discovery classification, and quantum-chemistry prediction. To account for the wide variation in structural complexity across pre-training molecules, we further propose a $\textbf{chemistry-informed curriculum}$: five structural complexity descriptors partition the corpus into five tiers of increasing chemical difficulty, consistently outperforming non-curriculum baselines. We show that a single rendered bond-line image, processed by a vision encoder, is sufficient for competitive molecular property prediction, i.e. $\textit{chemical insight from sight alone}$. The best curriculum-trained configuration achieves the top result on $\textbf{5 of 10}$ benchmarks and top two on $\textbf{all 10}$, at $\textbf{$\textit{80$\times$ lower}$}$ FLOPs than the nearest multi-modal competitor.
Abstract:Every molecule ever synthesised can be drawn as a 2D skeletal diagram, yet in modern property prediction this universally available representation has received less focus in favour of molecular graphs, 3D conformers, or billion-parameter language models, each imposing its own computational and data-engineering overhead. We present $\textbf{MolSight}$, the first systematic large-scale study of vision-based Molecular Property Prediction (MPP). Using 10 vision architectures, 7 pre-training strategies, and $2\,M$ molecule images, we evaluate performance across 10 downstream tasks spanning physical-property regression, drug-discovery classification, and quantum-chemistry prediction. To account for the wide variation in structural complexity across pre-training molecules, we further propose a $\textbf{chemistry-informed curriculum}$: five structural complexity descriptors partition the corpus into five tiers of increasing chemical difficulty, consistently outperforming non-curriculum baselines. We show that a single rendered bond-line image, processed by a vision encoder, is sufficient for competitive molecular property prediction, i.e. $\textit{chemical insight from sight alone}$. The best curriculum-trained configuration achieves the top result on $\textbf{5 of 10}$ benchmarks and top two on $\textbf{all 10}$, at $\textbf{$\textit{80$\times$ lower}$}$ FLOPs than the nearest multi-modal competitor.
Computer Vision and Pattern Recognition (cs.CV); Computation and Language (cs.CL)
arXiv:2605.10157 [cs.CV]
(or arXiv:2605.10157v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2605.10157 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 273 cuvinte
Subiecte conexe
MolEmb: Multimodal Large Language Models Can Be Strong Molecular Embedding Models
arXiv · 27 aug. 2026
Quantum scientists map exciton wave functions in organic films
Quantum Zeitgeist · 1 sept. 2026
A General-Purpose Molecular Foundation Model Transfers Across Diverse Olfactory Tasks
arXiv · 27 aug. 2026
A crystal’s natural shape focuses light like a waveguide
Quantum Zeitgeist · 12 aug. 2026