Research map / Probabilistic and scientific modeling / Molecules and proteins
ANTN: Bridging Autoregressive Neural Networks and Tensor Networks for Quantum Many-Body Simulation
Zhuo Chen, Laker Newhouse, Eddie Chen, Di Luo, Marin Soljacic
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 autoregressive neural network on top of MPS (elementwise ANTN, bond dim 8) (ablation) | MPS (8) · J1-J2 10x10, J2=0.2 | energy per site: -1.997537 -> -2.26034(6) | Improved |
| add autoregressive neural network on top of MPS (blockwise ANTN, bond dim 70) (ablation) | MPS (70) · J1-J2 10x10, J2=0.2 | energy per site: -2.191048 -> -2.25755(8) | Improved |
| add MPS structure to PixelCNN (elementwise ANTN) (ablation) | PixelCNN · J1-J2 10x10, J2=0.5 | energy per site: -1.74098(29) -> -1.94001(6) | Improved |
| add MPS structure to PixelCNN (blockwise ANTN) (ablation) | PixelCNN · J1-J2 12x12, J2=0.8 | energy per site: -2.03226(59) -> -2.09113(43) | Improved |
| elementwise ANTN (bond dim 8) versus MPS with bond dim 1024 on a small system (ablation) | Elementwise ANTN (8) · J1-J2 10x10, J2=0.6 | energy per site: -1.865561 vs -1.85684(10) | Hurt |
| increase system size from 10x10 to 12x12 when comparing elementwise ANTN (8) with MPS (1024) (ablation) | Elementwise ANTN (8) · J1-J2, J2=0.8 | energy per site: -2.062730 vs -2.05707(43) at 10x10; -2.093013 vs -2.09457(10) at 12x12 | Improved |
| increase system size from 10x10 to 12x12 when comparing elementwise ANTN (8) with MPS (1024) (ablation) | Elementwise ANTN (8) · J1-J2, J2=0.6 | energy per site: -1.865561 vs -1.85684(10) at 10x10; -1.858846 vs -1.85443(15) at 12x12 | Mixed |
| use ANTN instead of matrix product backflow state with sign rule (ablation) | Elementwise ANTN (8) · J1-J2 10x10, J2=0.5 | energy per site: -1.9267 -> -1.94001(6) | Improved |
| apply Marshall sign rule to RBM and PixelCNN (ablation) | RBM / PixelCNN · J1-J2, J2=0.2, 0.5, 0.8 | energy per site | Mixed |
| remove the need for the sign rule in ANTN (ablation) | ANTN · J1-J2 Heisenberg | energy per site | No clear effect |
| provide the sign rule to MPS on shallow random circuit states (ablation) | MPS · 16-qubit shallow random circuit states | fidelity | No clear effect |
| use ANTN with bond dim 4 versus MPS with bond dim 16 (ablation) | ANTN · 16-qubit shallow random circuit states | fidelity: bond dimension 4 vs 16 | 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.
- PixelCNN — PixelCNN++: Improving the PixelCNN with Discretized Logistic Mixture Likelihood and Other Modifications (ICLR 2017) · better on 7 of 7 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
- One More Time: Revisiting Neural Quantum States from a Reinforcement Learning Perspective (NeurIPS 2026 · similarity 0.88)
- Expressive power of tensor-network factorizations for probabilistic modeling (NeurIPS 2019 · similarity 0.88)
- Diagonal Symmetrization of Neural Network Solvers for the Many-Electron Schrödinger Equation (ICML 2025 · similarity 0.86)
- TensorRL-QAS: Reinforcement learning with tensor networks for improved quantum architecture search (NeurIPS 2025 · similarity 0.86)
- Predicting Ground State Properties: Constant Sample Complexity and Deep Learning Algorithms (NeurIPS 2024 · similarity 0.86)
- PALQO: Physics-informed model for Accelerating Large-scale Quantum Optimization (NeurIPS 2025 · similarity 0.84)
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.
