A new arXiv paper argues that evaluating quantum architecture search only by energy accuracy hides circuit cost and the physical properties of the prepared state. The authors present HamQASBench, a structure-aware benchmark and evaluation protocol aimed at molecular ground-state estimation. It is a methodology and benchmarking contribution rather than a model release.
DreamQAS targets reinforcement-learning-based quantum architecture search, where a variational quantum eigensolver is run again after every added gate even though the circuit transitions and legal actions are already known. The method keeps those exact dynamics and instead learns a decision-useful world model, reducing the number of VQE evaluations needed during search. It is a research preprint posted to arXiv.
Reinforcement learning can automate the design of quantum circuits, but performance degrades as the search space grows and rewarding circuit trajectories become rare. This arXiv paper proposes a generative replay technique that synthesizes new training samples instead of only reusing previously observed transitions, addressing the resulting data shortage. The authors report that this approach improves the scalability of quantum architecture search.