Research map / Probabilistic and scientific modeling / Molecules and proteins
QH9: A Quantum Hamiltonian Prediction Benchmark for QM9 Molecules
Haiyang Yu, Meng Liu, Youzhi Luo, Alex Strasser, Xiaofeng Qian, Xiaoning Qian, Shuiwang Ji
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 |
|---|---|---|---|
| train on QH9-stable-ood (OOD) instead of stable-id (ID) | QHNet · Intersection test set | Hamiltonian MAE all: 57.53 -> 72.78 | Hurt |
| train on QH9-stable-ood instead of stable-id | QHNet · Intersection test set | eigenenergy MAE: 442.78 -> 630.49 | Hurt |
| train on QH9-stable-ood instead of stable-id | QHNet · Intersection test set | orbital coefficient cosine similarity: 95.26 -> 94.01 | Hurt |
| OOD training on 29-atom molecules | QHNet · Intersection test set, 29 atoms | Hamiltonian MAE all: 47.00 -> 96.15 | Hurt |
| molecule-wise instead of geometry-wise split (100k) | QHNet · QH9-dynamic-100k | Hamiltonian MAE all: 96.85 -> 173.92 | Hurt |
| molecule-wise instead of geometry-wise split (300k) | QHNet · QH9-dynamic-300k | Hamiltonian MAE all: 100.19 -> 119.66 | Hurt |
| more trajectories: 300k-mol instead of 100k-mol | QHNet · QH9-dynamic-mol | Hamiltonian MAE all: 173.92 -> 119.66 | Improved |
| more trajectories: 300k-mol instead of 100k-mol | QHNet · QH9-dynamic-mol | eigenenergy MAE: 9719.58 -> 2178.15 | Improved |
| more trajectories: 300k-geo instead of 100k-geo | QHNet · QH9-dynamic-geo | Hamiltonian MAE all: 96.85 -> 100.19 | No clear effect |
| reduce training set from 104,664 to 10,000 (ablation) | QHNet · QH9-stable-id | eigenenergy MAE: 798.51 -> 2471.46 | Hurt |
| reduce training set from 104,664 to 10,000 (ablation) | QHNet · QH9-stable-id | Hamiltonian MAE all: 76.31 -> 78.60 | No clear effect |
| reduce training set from 104,664 to 10,000 (ablation) | QHNet · QH9-stable-id | orbital coefficient cosine similarity: 95.85 -> 96.18 | No clear effect |
Closest papers by meaning
- $\nabla^2$DFT: A Universal Quantum Chemistry Dataset of Drug-Like Molecules and a Benchmark for Neural Network Potentials (NeurIPS 2024 · similarity 0.89)
- Enhancing the Scalability and Applicability of Kohn-Sham Hamiltonians for Molecular Systems (ICLR 2025 · similarity 0.88)
- High-order Equivariant Flow Matching for Density Functional Theory Hamiltonian Prediction (NeurIPS 2025 · similarity 0.88)
- Learning the Electronic Hamiltonian of Large Atomic Structures (ICML 2025 · similarity 0.87)
- Self-Consistency Training for Density-Functional-Theory Hamiltonian Prediction (ICML 2024 · similarity 0.87)
- Infusing Self-Consistency into Density Functional Theory Hamiltonian Prediction via Deep Equilibrium Models (NeurIPS 2024 · similarity 0.86)
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.
