Research map / Graph learning / Graph self-supervised learning
GeoTMI: Predicting Quantum Chemical Property with Easy-to-Obtain Geometry via Positional Denoising
Hyeonsu Kim, Jeheon Woo, SEONGHWAN KIM, Seokhyun Moon, Jun Hyeong Kim, Woo Youn Kim
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add GeoTMI training (3-term MI maximization with denoising) to EGNN trained on MMFF geometry | EGNN · QM9 R2 | MAE: 5.60 -> 4.08 | Improved |
| add GeoTMI to DimeRxn trained on reactant/product geometry | DimeRxn · CCSD(T)-UNI | barrier height MAE: 6.03 -> 3.90 | Improved |
| add GeoTMI to DimeRxn trained on reactant/product geometry | DimeRxn · B97-D3 | barrier height MAE: 7.32 -> 4.17 | Improved |
| add GeoTMI on top of Noisy Nodes to Equiformer* (4 blocks) | Equiformer* · OC20 IS2RE average | energy MAE: 0.513 -> 0.492 | Improved |
| add Noisy Nodes to Equiformer* | Equiformer* · OC20 IS2RE average | EwT: 3.66 -> 4.72 | Improved |
| remove denoising loss L_d (ablation) | GeoTMI (EGNN) · QM9 U0 | MAE: 21.0 -> 14.5 | Improved |
| add position update (ablation) | GeoTMI (DimeRxn) · CCSD(T)-UNI | barrier height MAE: 5.77 -> 3.90 | Improved |
| add position update (ablation) | GeoTMI (DimeRxn) · B97-D3 | barrier height MAE: 7.19 -> 4.17 | Improved |
| add position update (ablation) | GeoTMI (EGNN) · QM9 U0 | MAE: 14.2 -> 14.5 | No clear effect |
| add L_y,correct (ablation) | GeoTMI (EGNN) · QM9 Cv | MAE: 0.0401 -> 0.0407 | No clear effect |
| use gradual denoising instead of restoring X at every layer (ablation) | GeoTMI (EGNN) · QM9 U0 | MAE: 15.4 -> 14.5 | Improved |
| use gradual denoising instead of denoising last layer only (ablation) | GeoTMI (EGNN) · QM9 mu | MAE: 0.132 -> 0.0997 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Equiformer* — Equiformer: Equivariant Graph Attention Transformer for 3D Atomistic Graphs (ICLR 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Simple GNN Regularisation for 3D Molecular Property Prediction and Beyond (ICLR 2022 · similarity 0.87)
- 3D Infomax improves GNNs for Molecular Property Prediction (ICML 2022 · similarity 0.87)
- Learning the Electronic Hamiltonian of Large Atomic Structures (ICML 2025 · similarity 0.87)
- GeoMFormer: A General Architecture for Geometric Molecular Representation Learning (ICML 2024 · similarity 0.86)
- Pushing the Limits of All-Atom Geometric Graph Neural Networks: Pre-Training, Scaling, and Zero-Shot Transfer (ICLR 2025 · similarity 0.85)
- Efficient and Scalable Density Functional Theory Hamiltonian Prediction through Adaptive Sparsity (ICML 2025 · similarity 0.85)
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.
