AI Paper Atlas

Research map / Graph learning / Graph self-supervised learning

Learning the Neighborhood: Contrast-Free Multimodal Self-Supervised Molecular Graph Pretraining

Boshra Ariguib, Mathias Niepert, Andrei Manolache

ICML 2026 · Graph learning › Graph self-supervised learning · approach: Molecular graph pretraining

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add DeepSets subgraph aggregation to the 2D-only probeC-FREE 2D · MoleculeNet linear probeROC-AUC avg: 66.63 -> 67.27Improved
add 3D conformers to the 2D backbone (multimodal, whole-molecule probe)C-FREE · MoleculeNet linear probeROC-AUC avg: 66.63 -> 71.07Improved
add DeepSets subgraph aggregation to the multimodal probeC-FREE MM · MoleculeNet linear probeROC-AUC avg: 71.07 -> 70.92No clear effect
use PaiNN encoder with three conformers instead of SchNet single conformerC-FREE · MoleculeNet full fine-tuningROC-AUC avg: 76.63 -> 79.81Improved
add pretraining to PaiNN on QM9PaiNN · QM9 AlphaMAE: 0.149 -> 0.107Improved
add pretraining to PaiNN on QM9 MuPaiNN · QM9 MuMAE: 0.042 -> 0.057Hurt
add chemically informed Murcko targets instead of EgoNetsC-FREE 3D · QM9MAE: 1.155 -> 1.135Mixed
add self-supervised pretraining at 1% labelsDrugs-75K model · Drugs-75K IPMAE: 0.638 -> 0.608Improved
add self-supervised pretraining at 100% labelsDrugs-75K model · Drugs-75K IPMAE: 0.419 -> 0.419No clear effect
add 2D modality to 3D-only backbone (ablation)C-FREE fine-tuned · Kraken LMAE: 0.329 -> 0.306Improved
add a transformer predictor (vs none) (ablation)C-FREE 2D · Kraken B5 linear probeMAE: 1.065 -> 0.588Improved
add a transformer predictor (vs MLP) (ablation)C-FREE 2D · Kraken B5 linear probeMAE: 0.817 -> 0.588Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy