Research map / Architectures and efficiency / RNNs and normalization
WF-Bench: A Benchmark for Neural-Network WaveFunction Expressivity and Scaling Laws
Lixing Zhang, Guijing Duan, Di Luo
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 |
|---|---|---|---|
| Use Psiformer instead of Ferminet at comparable parameter count | NN wavefunction architecture · WF-Bench, all 31 target wavefunctions, F(8) | fidelity: Laughlin m5: 0.8737 -> 0.9966 | Improved |
| Increase number of determinants Ndet beyond about 32 (ablation) | Ferminet and Psiformer · bcs s and moore m2 | fidelity | No clear effect |
| Increase number of determinants Ndet at small values (ablation) | Ferminet and Psiformer · bcs s and moore m2 | fidelity | Improved |
| Increase number of layers from 1 to 2 (ablation) | Ferminet and Psiformer · bcs s and moore m2 | fidelity | Improved |
| Increase number of layers beyond 2 (ablation) | Ferminet and Psiformer · bcs s and moore m2 | fidelity | No clear effect |
| Use only one layer for the network (ablation) | Ferminet and Psiformer · moore m2 | fidelity | Hurt |
| Increase electron number Ne | Ferminet and Psiformer · all WF-Bench wavefunctions | fidelity | Hurt |
| Target a chiral triplet superconducting state (bcs fp tp) instead of a singlet state | Ferminet and Psiformer · BCS wavefunctions | fidelity decay: Ferminet F(8): 0.9925 (bcs s) -> 0.8902 (bcs fp tp) | Hurt |
| Add quasiholes to the Laughlin m=3 state | Ferminet and Psiformer · topological states | fidelity | Hurt |
| Mix target and network distributions in the probability-matching loss (pmix) as pretraining | topological state training · WF-Bench topological states | training stability | Improved |
Closest papers by meaning
- One More Time: Revisiting Neural Quantum States from a Reinforcement Learning Perspective (NeurIPS 2026 · similarity 0.85)
- Learning the Electronic Hamiltonian of Large Atomic Structures (ICML 2025 · similarity 0.84)
- Enhancing the Scalability and Applicability of Kohn-Sham Hamiltonians for Molecular Systems (ICLR 2025 · similarity 0.83)
- Predicting Ground State Properties: Constant Sample Complexity and Deep Learning Algorithms (NeurIPS 2024 · similarity 0.83)
- Neural Pfaffians: Solving Many Many-Electron Schrödinger Equations (NeurIPS 2024 · similarity 0.83)
- The Importance of Being Scalable: Improving the Speed and Accuracy of Neural Network Interatomic Potentials Across Chemical Domains (NeurIPS 2024 · similarity 0.82)
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.
