Chengyong Yu | arXiv Science

arXiv2026

Emergent Incompressibility in Polariton Condensates for Scalable XY Spin Simulation

Driven-dissipative polariton networks offer a powerful analog platform for simulating continuous-spin Hamiltonians. However, their practical scalability is fundamentally limited by amplitude-phase coupling and metastable trapping in highly frustrated energy landscapes. Here, we present an \textit{in situ} dual-enhancement mechanism to address these bottlenecks without requiring external site-specific fine-tuning. First, a renormalization-group analysis based on density rescaling shows that higher-order nonlinear loss drives the condensates toward an incompressible regime, intrinsically suppressing amplitude fluctuations and enforcing phase-only dynamics. For macroscopic XY networks (up to 10,000 spins), this phase-amplitude decoupling yields a more than 50-fold reduction in energy, while simultaneously shrinking the standard deviation by a factor of six. More generally, it restores the extensive scaling of the optimized XY energy with system size. Second, to navigate the rugged, nonconvex energy landscape, we introduce weak Hermitian pulses. When implemented through a multistage protocol, these pulses enable repeated escapes from local minima and drive the system to deeper energy minima. Benchmarking shows an increasing advantage with system size, with the polariton simulator outperforming the classical Monte Carlo-SQP baseline in every sampled macroscopic realization. Our work provides a physical framework for scalable polariton analog computing.

cond-mat.mes-hall↗

arXiv2026

Robust Quantum Learning through Hamiltonian Reservoir Computing

Quantum learning provides a versatile paradigm for information processing by exploiting the intrinsic representational capacity of high-dimensional Hilbert spaces. Here, we investigate a Hamiltonian-encoding framework for quantum reservoir computing that simultaneously addresses three key challenges in quantum learning: trainability, hardware efficiency, and information stability. In this framework, input data are directly mapped onto a fixed Hamiltonian and transformed into expressive nonlinear features through quantum dynamical evolution. By employing the reservoir-computing paradigm, the approach naturally circumvents the barren plateau problem in quantum learning landscapes. We validate the framework across two complementary platforms: an analog superconducting array processor and a digital gate-based quantum circuit implementation. Despite their fundamentally different realizations, both platforms exhibit comparable representational power and achieve competitive learning performance, establishing a unified framework for cross-platform quantum learning. While both implementations achieve comparable performance, the analog processor may offer a more hardware-efficient realization by bypassing the temporal overhead of gate-based decomposition and thereby making more effective use of finite coherence times, albeit at the expense of universality. Furthermore, we find that finite dissipation suppresses quantum-scrambling-induced instabilities at long evolution times and can enhance learning performance, revealing a constructive role for environmental coupling in stabilizing quantum learning dynamics. Collectively, these results establish Hamiltonian-encoded reservoir computing as a compact, expressive, and hardware-efficient paradigm for quantum learning on current-generation quantum platforms.

quant-ph↗