⚛️ Physics

🧪 ICML2026 · 1 paper notes

📌 Same area in other venues: 📷 CVPR2026 (1) · 🔬 ICLR2026 (2) · 🤖 AAAI2026 (2) · 🧠 NeurIPS2025 (18) · 📹 ICCV2025 (1)

Neural QAOA\(^2\): Differentiable Joint Graph Partitioning and Parameter Initialization for Quantum Combinatorial Optimization

A generative-evaluative neural network (GEN) jointly differentiates the two tasks of "graph partitioning + quantum circuit parameter initialization" in QAOA²: the evaluator learns a high-fidelity quantum performance surrogate, and the generator, guided by its gradients, outputs discrete partitions and parameter initializations. With straight-through estimator (STE) and orthogonal complement head (OCH), the system is end-to-end trainable. Outperforms heuristic baselines on 183 QUBO/Ising/MaxCut instances (21-1000 variables), ranking first on 101 cases.