These results establish a world-model design for QAS whose value lies in decision-useful feedback rather than exact energy prediction, and establish a world-model design for QAS whose value lies in decision-useful feedback rather than exact energy prediction.
Abstract
Reinforcement-learning-based quantum architecture search (RL-QAS) repeatedly optimizes a variational quantum eigensolver (VQE) after extending a circuit, although circuit construction and action legality are deterministic and known. We introduce DreamQAS, a model-based RL framework that preserves these exact circuit dynamics and learns only the expensive post-VQE feedback. A recurrent randomized-prior ensemble predicts an oracle-free score relative to an empirical energy frontier and supports multi-step imagined policy learning over explicit legal circuits. Ranking-based activation, uncertainty-aware pessimism and truncation, and selective real-VQE verification form a reliability-controlled learning loop. Under a common 15,000-episode budget and frozen evaluation for the RL methods, DreamQAS has the lowest mean frozen-policy energy error on four of five molecular tasks and the second-lowest on one. At fine-error targets reached by all seeds of both methods, it uses 1.6x to 2.0x fewer real VQE calls on four tasks and 10.6x fewer on BeH2-8q. Counterfactual action-ranking utility increases across all five tasks, with a mean increase of 0.346 and a 95 percent confidence interval of [0.185, 0.507], while direct greedy and beam use of the same model does not recover the gains of imagined policy learning. Ensemble disagreement also improves risk-coverage over random rejection on all three probed tasks. These results establish a world-model design for QAS whose value lies in decision-useful feedback rather than exact energy prediction.
Quantum reservoir computing (QRC) uses fixed quantum dynamics as a high-dimensional temporal feature map and trains only a lightweight classical readout. QRC is attractive for near-term quantum machine learning, but its performance depends strongly on architecture choices such as input encoding, reservoir depth, entanglement topology, measurement features, state-reset policy, feature construction, and readout regularization. We introduce \method, a simulator-based benchmark that formulates QRC design as constrained black-box architecture search and evaluates whether large language models can act as proposal controllers for this search problem. The benchmark compares five policies under identical evaluation budgets: random search, evolutionary search, Bayesian/TPE optimization, a feedback-based LLM agent, and \hybrid, which combines LLM proposals with memory, mutation, crossover, duplicate avoidance, and exploration. On NARMA10, Mackey-Glass forecasting, and temporal parity, \hybrid{} is the most consistent policy: it ranks first on NARMA10 and temporal parity and second on Mackey-Glass, narrowly behind evolutionary search. Under a 25-evaluation budget and three seeds, \hybrid{} improves over random search on all tasks, including a 23.6\% relative reduction in Mackey-Glass error. The results do not show that LLMs are universal QRC optimizers; rather, they show that generative models can be useful high-level controllers when embedded inside validated, reproducible hybrid search loops.
Building a quantum machine learning (QML) model competitive with a classical baseline currently requires a practitioner to separately choose a circuit architecture, a data-encoding scheme, a model paradigm (kernel versus variational), and a set of training hyperparameters, then verify after the fact that the chosen circuit is even trainable. Existing QML libraries provide the primitives for this but not the search, and existing classical AutoML libraries provide the search but not the quantum-specific search space or diagnostics. We present qkabrine-automl, a Python package that treats architecture, encoding, model type, and hyperparameters as a single, jointly searchable configuration space, evaluated through one consistent harness regardless of which of five search strategies proposed the candidate. The package integrates trainability diagnostics, a Data Quantum Fisher Information Metric (DQFIM) estimate and a gradientmagnitude barren-plateau monitor, directly into the evaluation loop as an optional prescreening step, alongside expressibility and entangling-capability characterization, a post-search circuitsurgery pass for NISQ deployment, and OpenQASM export. We position this contribution against recent AutoQML frameworks that already automate parts of the QML pipeline, and report a small, fully reproducible illustrative run rather than a benchmark claim.
RubriQ is introduced, a scalable framework that formulates circuit synthesis as a large language model (LLM) code-generation task, optimized via group relative policy optimization (GRPO), which establishes an automated, high-performance computing (HPC)-driven pipeline for generating hardware-ready, fault-tolerant quantum circuits at scale.
Numerical results show that the learned architectures recover known benchmark strategies, adapt to dephasing noise, and outperform fixed hardware-efficient ans\"atze while using fewer entangling gates, establishing AutoQSense as a resource-aware approach to adaptive and hardware-compatible quantum sensing.
Sample-based quantum diagonalization (SQD), equivalently quantum-selected configuration interaction (QSCI), has in two years become a pragmatic centre of gravity of pre-fault-tolerant quantum chemistry: a quantum processor samples electronic configurations, and the many-electron Hamiltonian is diagonalized classically in the resulting determinant subspace. Its accuracy is set entirely by which configurations enter that subspace, a selection problem for machine learning made acute by a coupon-collector bottleneck. We critically review the ecosystem of generative and learned selectors, organizing it by the object each method generates and the importance signal it exploits, and expose one conspicuous gap: a reward-proportional generative-flow-network proposer built for tail discovery. We then confront the field's central question -- whether the quantum sampler beats classical selected configuration interaction -- and report a carefully scoped negative: across published same-active-space comparisons, strong classical selected CI matches or beats the quantum-sampled subspace, and the flagship single-layer circuits now admit polynomial-time classical energy estimation. We distil a benchmarking standard and turn the negative into a regime map, then test it with FCI-exact experiments that confirm one prediction and refute another: the cheap prior's rank correlation with the exact weights declines with multireference character (a usable coordinate), but a controlled single-molecule noise sweep shows the one generative advantage we find, robustness to valid-shot starvation, to be generic rather than the multireference-specific effect a confounded contrast first suggested. Finally, we flag learning from quantum experiments, whose classical sample-complexity lower bound is an unconditional theorem, as the one adjacent frontier where a quantum advantage is provable but not yet bridged to chemistry.
A weeklong summer workshop brought higher education faculty to campus to explore how AI and machine learning materials can be adapted for their classrooms.
