Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

Journal of Quantum Artificial Intelligence

Causal Representation Learning with Quantum Encodings for Personalized Decision Policies in Adaptive Systems

Aldy Agil Hermanto

Published Aug 31, 2026• 0 citations
Publisher

Abstract

Personalized adaptive systems frequently learn from biased logs generated by historical policies, where spurious correlations and environment shift degrade both generalization and cohort safety. This paper introduces a unified pipeline that couples causal representation learning with quantum encodings to produce stable embeddings for offline-trained decision policies. Across environment-split evaluations, the proposed quantum causal representation reduced the invariance gap from 0.118 (baseline classical) and 0.067 (causal classical) to 0.041, while improving reward prediction AUROC from 0.791 and 0.816 to 0.842. Under drifted splits, representation shift-drop decreased from 7.6% to 2.5%, indicating stronger transfer under distributional change. Policies trained on these embeddings achieved higher normalized off-policy value, improving from 0.456 (baseline) and 0.472 (causal classical) to 0.489, while reducing shift gap from 6.9% to 2.6% and increasing worst-cohort uplift to 0.028. Drift stress tests showed sustained value retention at severe drift (level 3), with values 0.401 (baseline), 0.433 (causal classical), and 0.459 (quantum causal), accompanied by lower calibration error (0.084, 0.068, and 0.056, respectively). Ablations demonstrated a capacity sweet spot at 6 qubits and 4 layers, delivering strong value (0.489) at deployable latency (6.7 ms), whereas higher capacity yielded marginal value gains (0.491) at increased latency (9.4 ms). These results indicate that integrating causal invariance with carefully budgeted quantum feature maps improves personalization effectiveness, robustness, and reliability in adaptive decision-making from logs.

DOI
10.63913/jqai.v1i2.10

Related papers

Context-Aware Error Mitigation Orchestration for Hybrid Quantum Reinforcement Learning on NISQ Systems

Bisma Majid, Shabir Ahmed Sofi, Mir Mohammad Yousuf

Quantum Reinforcement Learning (QRL) integrates reinforcement learning with parameterized quantum circuits and is a promising approach to combinatorial optimization. On Noisy Intermediate-Scale Quantum (NISQ) devices, however, decoherence, gate imperfections, and measurement errors reduce policy quality and make learning less reliable. Existing error mitigation techniques are generally applied as fixed corrections that do not adapt to changing noise conditions or to the evolving state of training. This work presents Adaptive Policy-Guided Error Mitigation (APGEM) as a context-aware orchestration layer of the hybrid quantum-classical training loop that dynamically selects the most suitable mitigation strategy during QRL training. APGEM evaluates Zero-Noise Extrapolation (ZNE), Probabilistic Error Cancellation (PEC), Clifford Data Regression (CDR), and Readout Error Mitigation (REM) using policy-level indicators, including quantum-state fidelity, policy entropy, cumulative reward, and approximation ratio, and integrates the selected strategy directly into the reinforcement learning loop. The framework is evaluated on the Capacitated Vehicle Routing Problem (CVRP), a representative NP-hard problem in urban logistics, under a range of NISQ noise models and noise levels. APGEM consistently outperforms conventional static mitigation methods, reaches approximately 94% of the utility of an oracle strategy, maintains higher quantum-state fidelity as noise increases, and produces more stable learning behaviour throughout training. Ablation studies show that the framework learns context-aware mitigation policies that adapt to different noise environments and circuit execution conditions. These findings demonstrate that integrating adaptive error mitigation into the learning process substantially improves the robustness and reliability of QRL on NISQ hardware.

cs.LGCross-listedarXiv:2610.01253•Oct 2, 2026•0 citations

Quantum-Compatible Causal Policy Learning: A Vision for Constrained Personalized Treatment Decisions

Deepika Gopukumar, Abhijeet Ghoshal, Nirup Menon +2 more

Personalized treatment recommendation requires translating heterogeneous causal evidence into an interpretable policy while balancing expected benefit, individual regret, and consistency across clinically similar patients. In this position paper, we argue that quantum optimization is justified in this setting only if the downstream policy-optimization problem is hard for strong classical methods, small and structured enough for quantum hardware, and consequential for patient outcomes. We build on an established causal policy-learning pipeline in which classical causal machine learning methods first provide patient-specific doubly robust treatment rewards, and we focus on the optimization layer that follows. A small proof of concept using five clinical covariates and ten patients shows that progressively richer policy decisions, including rule selection, treatment assignment, candidate-structure selection, and hard local-consistency constraints, can be encoded as Quadratic Unconstrained Binary Optimization (QUBO) problems whose evaluation reproduces the corresponding classical optimum. These checks establish formulation feasibility over the evaluated configurations, a precondition for any quantum experiment. Because exact search over fixed-depth policy trees runs in polynomial time, we locate the open question in constraints that couple patients across tree leaves and set out four research questions, each with a stop rule.

Proceedings of the AAAI Symposium Series•Nov 6, 2026•0 citations

Quantum-Enhanced Reinforcement Learning for Adaptive Decision-Making in Dynamic Environments

Priyanka Singh, Sonam Dubey

Many challenges that occur with traditional reinforcement learning in uncertain environments, such as slow convergence and scalability, can be addressed with Quantum-Enhanced Reinforcement Learning (QERL). Specifically, this research is for adaptive decision-making in environments with rapid change and great uncertainty. A QERL framework that combines variational quantum circuits and deep Q-networks is introduced. This framework optimizes decision-making processes in real-time and is applicable in areas such as autonomous systems, smart manufacturing, and intelligent resource allocation. Training for the framework took place in simulated dynamic environments during 50,000 training episodes and was then evaluated alongside Deep Q-Learning and Proximal Policy Optimization models. The QERL self-model outperformed the baseline in policy accuracy (+18.6%) and cumulative reward loss (-21.4%). Decision latency decreased 16.2%, and convergence improved 23.8% with the use of a hybrid quantum architecture. The proposed analysis demonstrated improved average reward (78.4 to 94.7) with 95% CI, indicating model consistency. Ultimately, QERL improves decision-making in complex scenarios and provides robust adaptive learning. Also noted, QERL frameworks with hybrid quantum architecture are useful for real-world systems. Furthermore, the learning framework proved to be more stable and adaptable to rapid and stochastic environmental changes compared to classical agents, in environments with incomplete dynamic and rapidly changing environments.

•Aug 1, 2026•0 citations

Self-Supervised Contrastive Learning in Quantum Latent Spaces for Few-Shot Domain Adaptation

Christianto Hernando

Few-shot domain adaptation remains fragile under compound distribution shifts because sparse target supervision amplifies representation noise and destabilizes metric structure. This study introduces self-supervised contrastive learning directly in a quantum latent space produced by a variational quantum circuit encoder, followed by prototype-driven adaptation with distribution alignment. Across three target shifts (style, temporal sparsity, and mixed covariate plus conditional drift), the proposed method achieved mean episodic accuracies of 0.721, 0.705, and 0.689 at 1-shot, increasing to 0.832, 0.813, and 0.806 at 5-shot, and 0.851, 0.836, and 0.826 at 10-shot, respectively. At K=5, average accuracy improved from 0.672 under ERM to 0.824 with the full quantum-contrastive model, with 95% confidence intervals tightening from ±0.018 to ±0.011. Representation diagnostics showed a reduction in intra-class dispersion from 0.412 to 0.301 and an increase in inter-class separation from 0.688 to 0.824, while seed variability decreased from 0.012 to 0.007. Circuit ablations identified an optimal depth range of 6–8 layers, where accuracy peaked at 0.824–0.829 before declining at higher depth with increased episodic variance. Robustness tests under finite-shot measurement revealed graceful degradation: under severe shift, accuracy rose from 0.709 at 64 shots per observable to 0.801 at 2048 shots, with failure episodes reduced from 4.2% to 2.9% when increasing from 256 to 1024 shots. These results indicate that quantum-measurement-bounded embeddings can stabilize contrastive geometry and improve few-shot transfer when target labels are scarce.

Journal of Quantum Artificial Intelligence•Oct 17, 2026•0 citations

Quantum Representation Learning for Reinforcement Learning in Robot Navigation

Van Le, Le Thanh Tan, K. C. Ho

Reinforcement learning for robot navigation requires representations that capture spatial geometry, long-horizon dependencies, and goal-conditioned structure. Classical neural networks provide flexible function approximators, but their representations are ultimately limited by classical feature transformations. We investigate quantum representation learning as a mechanism for encoding state–goal pairs in goal-conditioned reinforcement learning. We formulate navigation as a finite-horizon Markov decision process and introduce a quantum variational feature map that embeds continuous states into a high-dimensional Hilbert space. This representation is optimized jointly with a Q-learning objective, yielding a hybrid quantum–classical value function. Classical networks serve only as baselines to validate the environment and evaluation protocol. After correcting subtle issues in reward shaping and success detection, we observe that quantum representations produce coherent value structures, stable optimization dynamics, and competitive success rates relative to classical baselines. Rather than aiming to improve navigation performance, our objective is to provide a controlled and reproducible framework for analyzing how quantum feature geometry shapes value estimation and goal-conditioned behavior in offline RL. Our analysis highlights how quantum feature geometry influences value estimation and navigation behavior, establishing a reproducible foundation for studying quantum-enhanced representations in reinforcement learning for robotics.

IEEE Access•Jan 1, 2026•0 citations

RDERP:Real-Time Decision-Making at the Edge Using Quantum-Assisted Reinforcement Learning and Predictive Optimization

Vonteru Srikanth Reddy, Patwari Krishna Rao, Mallu Shiva Rama Krishna

The surge in Internet of Things deployments and edge computing infrastructures has heightened the need for ultra low latency, high reliability decision-making systems. This paper presents RDERP: Real Time Decision Making at the Edge Using Quantum-Assisted Reinforcement Learning and Predictive Optimization, a hybrid quantum classical framework designed to enhance adaptability, scalability, processing efficiency in dynamic environments. The proposed system integrates quantum subroutines for accelerated policy evaluation with predictive optimization modules that forecast future system states using time series and probabilistic models. These forecasts enable proactive resource allocation, while quantum acceleration facilitates rapid convergence in large state–action spaces. The framework is evaluated using the Disaster Response Coordination dataset, which models real-time, multi-agent decision-making under uncertainty, along with additional datasets from smart manufacturing, healthcare monitoring, traffic optimization, and drone navigation. Performance is assessed across six key metrics: Decision Velocity, Real-Time Scalability Index, Energy Performance Trade-of Efficiency, Task Completion Assurance Rate, Multi-Agent Coordination Latency, Quantum Computational Gain. Experimental results demonstrate significant performance gains up to 38% reduction in decision latency, 27% faster policy adaptation, and 22% higher energy efficiency over baseline deep reinforcement learning models. Specifically, the DRC dataset achieved the highest DV of 25.9166 decisions/s, RSI of 0.928, and TCAR of 97.4192%, with a MACL of 3.08 ms. Quantum acceleration benefits are evidenced by QCG values ranging from 1.6525 to 1.8209. These results demonstrate that RDERP achieves a robust balance among speed, scalability, energy efficiency, and reliability, rendering it highly suitable for latency-sensitive and resource-constrained edge computing environments.

Procedia Computer Science•Jan 1, 2026•0 citations
Personalized adaptive systems frequently learn from biased logs generated by historical policies, where spurious correlations and environment shift degrade both generalization and cohort safety. This paper introduces a unified pipeline that couples causal representation learning with quantum encodings to produce stable embeddings for offline-trained decision policies. Across environment-split evaluations, the proposed quantum causal representation reduced the invariance gap from 0.118 (baseline classical) and 0.067 (causal classical) to 0.041, while improving reward prediction AUROC from 0.791 and 0.816 to 0.842. Under drifted splits, representation shift-drop decreased from 7.6% to 2.5%, indicating stronger transfer under distributional change. Policies trained on these embeddings achieved higher normalized off-policy value, improving from 0.456 (baseline) and 0.472 (causal classical) to 0.489, while reducing shift gap from 6.9% to 2.6% and increasing worst-cohort uplift to 0.028. Drift stress tests showed sustained value retention at severe drift (level 3), with values 0.401 (baseline), 0.433 (causal classical), and 0.459 (quantum causal), accompanied by lower calibration error (0.084, 0.068, and 0.056, respectively). Ablations demonstrated a capacity sweet spot at 6 qubits and 4 layers, delivering strong value (0.489) at deployable latency (6.7 ms), whereas higher capacity yielded marginal value gains (0.491) at increased latency (9.4 ms). These results indicate that integrating causal invariance with carefully budgeted quantum feature maps improves personalization effectiveness, robustness, and reliability in adaptive decision-making from logs.
Quantum Reinforcement Learning (QRL) integrates reinforcement learning with parameterized quantum circuits and is a promising approach to combinatorial optimization. On Noisy Intermediate-Scale Quantum (NISQ) devices, however, decoherence, gate imperfections, and measurement errors reduce policy quality and make learning less reliable. Existing error mitigation techniques are generally applied as fixed corrections that do not adapt to changing noise conditions or to the evolving state of training. This work presents Adaptive Policy-Guided Error Mitigation (APGEM) as a context-aware orchestration layer of the hybrid quantum-classical training loop that dynamically selects the most suitable mitigation strategy during QRL training. APGEM evaluates Zero-Noise Extrapolation (ZNE), Probabilistic Error Cancellation (PEC), Clifford Data Regression (CDR), and Readout Error Mitigation (REM) using policy-level indicators, including quantum-state fidelity, policy entropy, cumulative reward, and approximation ratio, and integrates the selected strategy directly into the reinforcement learning loop. The framework is evaluated on the Capacitated Vehicle Routing Problem (CVRP), a representative NP-hard problem in urban logistics, under a range of NISQ noise models and noise levels. APGEM consistently outperforms conventional static mitigation methods, reaches approximately 94% of the utility of an oracle strategy, maintains higher quantum-state fidelity as noise increases, and produces more stable learning behaviour throughout training. Ablation studies show that the framework learns context-aware mitigation policies that adapt to different noise environments and circuit execution conditions. These findings demonstrate that integrating adaptive error mitigation into the learning process substantially improves the robustness and reliability of QRL on NISQ hardware.
Personalized treatment recommendation requires translating heterogeneous causal evidence into an interpretable policy while balancing expected benefit, individual regret, and consistency across clinically similar patients. In this position paper, we argue that quantum optimization is justified in this setting only if the downstream policy-optimization problem is hard for strong classical methods, small and structured enough for quantum hardware, and consequential for patient outcomes. We build on an established causal policy-learning pipeline in which classical causal machine learning methods first provide patient-specific doubly robust treatment rewards, and we focus on the optimization layer that follows. A small proof of concept using five clinical covariates and ten patients shows that progressively richer policy decisions, including rule selection, treatment assignment, candidate-structure selection, and hard local-consistency constraints, can be encoded as Quadratic Unconstrained Binary Optimization (QUBO) problems whose evaluation reproduces the corresponding classical optimum. These checks establish formulation feasibility over the evaluated configurations, a precondition for any quantum experiment. Because exact search over fixed-depth policy trees runs in polynomial time, we locate the open question in constraints that couple patients across tree leaves and set out four research questions, each with a stop rule.
Many challenges that occur with traditional reinforcement learning in uncertain environments, such as slow convergence and scalability, can be addressed with Quantum-Enhanced Reinforcement Learning (QERL). Specifically, this research is for adaptive decision-making in environments with rapid change and great uncertainty. A QERL framework that combines variational quantum circuits and deep Q-networks is introduced. This framework optimizes decision-making processes in real-time and is applicable in areas such as autonomous systems, smart manufacturing, and intelligent resource allocation. Training for the framework took place in simulated dynamic environments during 50,000 training episodes and was then evaluated alongside Deep Q-Learning and Proximal Policy Optimization models. The QERL self-model outperformed the baseline in policy accuracy (+18.6%) and cumulative reward loss (-21.4%). Decision latency decreased 16.2%, and convergence improved 23.8% with the use of a hybrid quantum architecture. The proposed analysis demonstrated improved average reward (78.4 to 94.7) with 95% CI, indicating model consistency. Ultimately, QERL improves decision-making in complex scenarios and provides robust adaptive learning. Also noted, QERL frameworks with hybrid quantum architecture are useful for real-world systems. Furthermore, the learning framework proved to be more stable and adaptable to rapid and stochastic environmental changes compared to classical agents, in environments with incomplete dynamic and rapidly changing environments.
Few-shot domain adaptation remains fragile under compound distribution shifts because sparse target supervision amplifies representation noise and destabilizes metric structure. This study introduces self-supervised contrastive learning directly in a quantum latent space produced by a variational quantum circuit encoder, followed by prototype-driven adaptation with distribution alignment. Across three target shifts (style, temporal sparsity, and mixed covariate plus conditional drift), the proposed method achieved mean episodic accuracies of 0.721, 0.705, and 0.689 at 1-shot, increasing to 0.832, 0.813, and 0.806 at 5-shot, and 0.851, 0.836, and 0.826 at 10-shot, respectively. At K=5, average accuracy improved from 0.672 under ERM to 0.824 with the full quantum-contrastive model, with 95% confidence intervals tightening from ±0.018 to ±0.011. Representation diagnostics showed a reduction in intra-class dispersion from 0.412 to 0.301 and an increase in inter-class separation from 0.688 to 0.824, while seed variability decreased from 0.012 to 0.007. Circuit ablations identified an optimal depth range of 6–8 layers, where accuracy peaked at 0.824–0.829 before declining at higher depth with increased episodic variance. Robustness tests under finite-shot measurement revealed graceful degradation: under severe shift, accuracy rose from 0.709 at 64 shots per observable to 0.801 at 2048 shots, with failure episodes reduced from 4.2% to 2.9% when increasing from 256 to 1024 shots. These results indicate that quantum-measurement-bounded embeddings can stabilize contrastive geometry and improve few-shot transfer when target labels are scarce.
Reinforcement learning for robot navigation requires representations that capture spatial geometry, long-horizon dependencies, and goal-conditioned structure. Classical neural networks provide flexible function approximators, but their representations are ultimately limited by classical feature transformations. We investigate quantum representation learning as a mechanism for encoding state–goal pairs in goal-conditioned reinforcement learning. We formulate navigation as a finite-horizon Markov decision process and introduce a quantum variational feature map that embeds continuous states into a high-dimensional Hilbert space. This representation is optimized jointly with a Q-learning objective, yielding a hybrid quantum–classical value function. Classical networks serve only as baselines to validate the environment and evaluation protocol. After correcting subtle issues in reward shaping and success detection, we observe that quantum representations produce coherent value structures, stable optimization dynamics, and competitive success rates relative to classical baselines. Rather than aiming to improve navigation performance, our objective is to provide a controlled and reproducible framework for analyzing how quantum feature geometry shapes value estimation and goal-conditioned behavior in offline RL. Our analysis highlights how quantum feature geometry influences value estimation and navigation behavior, establishing a reproducible foundation for studying quantum-enhanced representations in reinforcement learning for robotics.
The surge in Internet of Things deployments and edge computing infrastructures has heightened the need for ultra low latency, high reliability decision-making systems. This paper presents RDERP: Real Time Decision Making at the Edge Using Quantum-Assisted Reinforcement Learning and Predictive Optimization, a hybrid quantum classical framework designed to enhance adaptability, scalability, processing efficiency in dynamic environments. The proposed system integrates quantum subroutines for accelerated policy evaluation with predictive optimization modules that forecast future system states using time series and probabilistic models. These forecasts enable proactive resource allocation, while quantum acceleration facilitates rapid convergence in large state–action spaces. The framework is evaluated using the Disaster Response Coordination dataset, which models real-time, multi-agent decision-making under uncertainty, along with additional datasets from smart manufacturing, healthcare monitoring, traffic optimization, and drone navigation. Performance is assessed across six key metrics: Decision Velocity, Real-Time Scalability Index, Energy Performance Trade-of Efficiency, Task Completion Assurance Rate, Multi-Agent Coordination Latency, Quantum Computational Gain. Experimental results demonstrate significant performance gains up to 38% reduction in decision latency, 27% faster policy adaptation, and 22% higher energy efficiency over baseline deep reinforcement learning models. Specifically, the DRC dataset achieved the highest DV of 25.9166 decisions/s, RSI of 0.928, and TCAR of 97.4192%, with a MACL of 3.08 ms. Quantum acceleration benefits are evidenced by QCG values ranging from 1.6525 to 1.8209. These results demonstrate that RDERP achieves a robust balance among speed, scalability, energy efficiency, and reliability, rendering it highly suitable for latency-sensitive and resource-constrained edge computing environments.