Aug 2026· Physica Scripta· Vol 101, pp. 346001· 0 citations· 43 references
Physics
TL;DR
A novel policy network, quantum gated recurrent neural network (QGRNN), which incorporates quantum superposition encoding and cyclic entanglement structures to achieve more efficient state-space exploration and employs a quantum–classical hybrid residual design that effectively integrates quantum measurement outcomes with classical inputs, thereby strengthening the representational capacity for temporal decision-making features.
Abstract
Performance breakthroughs in dynamic temporal decision-making tasks have long been a core research objective in reinforcement learning. Yet traditional policy networks, including long short-term memory (LSTM) networks, gated recurrent unit (GRU) and feedforward neural network (FNN), have long been hampered by inherent bottlenecks: low reward efficiency and training instability, which renders them less capable of meeting the high-precision decision-making demands of complex scenarios. This paper introduces a novel policy network, quantum gated recurrent neural network (QGRNN), which incorporates quantum superposition encoding and cyclic entanglement structures to achieve more efficient state-space exploration. Furthermore, the network employs a quantum–classical hybrid residual design that effectively integrates quantum measurement outcomes with classical inputs, thereby strengthening the representational capacity for temporal decision-making features. Trained over 800 episodes via proximal policy optimization in CartPole-v1, the QGRNN achieves an average reward of 380.3 ± 128.6 in the last 100 episodes-significantly outperforming GRU (275.9 ± 140.7), LSTM (239.7 ± 131.8), and FNN (219.2 ± 109.5). As a 4-qubit proof-of-concept, it achieves 45% high-performance episodes (rewards >450) and demonstrates low-resource, high-performance characteristics for temporal decision-making.
Surprisingly, frozen recurrent networks support recall over hundreds of steps and achieve competitive control performance across diverse POMDPs, while using less than half the number of trainable parameters required by their fully trainable counterparts.
This work revisits Unitary recurrent networks (uRNNs) and proposes three different versions of uRNNs as drop-in replacements for recurrent PPO architectures, and demonstrates that the simple recurrence and the added degree of freedom from the phase of the complex representations enable significant gains over baselines...
Sathya Kamesh Bhethanabhotla, E. Gavves, André Biedenkapp· arXiv.org· 0 citations
Results indicate that recursive quantum feature transformations can improve stability and out-of-sample performance for compact hybrid quantum--classical temporal models.
M. Lee, Yen-Ku Liu, Samuel Yen-Chi Chen et al.· arXiv.org· 0 citations
Recurrent neural networks (RNNs) offer linear-time scaling with sequence length while requiring only constant memory, yet they struggle to capture long-range dependencies due to vanishing gradients and limited receptive fields. To address these limitations, we introduce a second-order recurrent model in which the stand...
Aleix Salvador-Pomarol, A. Montanari, E. K. Miller et al.· 0 citations
Training quantum neural networks (QNNs) on near-term hardware remains hampered by two compounding difficulties: the exponential vanishing of gradient variance known as the barren plateau, and the $\mathcal{O}(L \cdot 2^n)$ time and memory cost of differentiating through an $n$-qubit, $L$-layer circuit. We propose RLQ-G...
Training RNNs on delayed decision-making tasks with progressively increasing temporal demands shows that temporal and decision-related computations can emerge through multiple dynamical regimes, while maintaining structured low-dimensional representations and comparable behavioural performance, mirroring biological pri...
C. Jarne, Ryeongkyung Yoon, Tahra L. Eissa et al.· Journal of Computational Neu...· 0 citations
Related blog posts
MIT News · Artificial Intelligence· news.mit.eduNov 17, 2026
By studying the computational foundations of language learning, Associate Professor Jacob Andreas builds intelligent systems that can communicate effectively with humans.
MIT News · Artificial Intelligence· news.mit.eduNov 17, 2026
AI coding agents are increasingly help developers fix bugs and build features largely on their own, but every interaction with their models consumes tokens. In each session, an agent sends instructions, tool descriptions, and session history to its model. Most efforts to reduce this cost focus on individual tasks. Yet developers return to the same […] The post TokenEconomizer: Learning from Experience to Make AI Coding Agents More Efficient appeared first on Microsoft Research.
MIT News · Artificial Intelligence· news.mit.eduNov 9, 2026