arXiv subjects

Junhyeong An

Publications and source records attributed to Junhyeong An.

3 recordsLinked to original sources

arXiv2026

Beyond the Positive Partial Transpose Squared Conjecture: The Qutrit Case

Entanglement swapping is a fundamental operation in quantum repeaters for establishing entanglement between distant parties. The positive partial transpose (PPT) squared conjecture asks whether two PPT entangled links can generate terminal entanglement through entanglement swapping, or equivalently, whether the composition of two PPT maps is always entanglement breaking. Motivated by this conjecture, we investigate the map-composition problem beyond the PPT setting. For qutrit completely positive (CP) maps, we prove that the composition of any CP map whose Choi matrix is $1$-undistillable with any CP map whose Choi matrix has Schmidt number at most two is entanglement breaking in either order. Moreover, we show that the cone of $1$-undistillable CP maps is exactly the largest qutrit cone of CP maps whose composition with every CP map whose Choi matrix has Schmidt number at most two is entanglement breaking in both orders. Finally, although map composition captures only the standard maximally entangled outcome in entanglement swapping, we prove that any $1$-undistillable two-qutrit state and any state of Schmidt number at most two cannot generate terminal entanglement under an arbitrary selective measurement on the intermediate systems.

quant-ph↗

arXiv2026

Unlearning for One-Step Generative Models via Unbalanced Optimal Transport

Recent advances in one-step generative frameworks, such as flow map models, have significantly improved the efficiency of image generation by learning direct noise-to-data mappings in a single forward pass. However, machine unlearning for ensuring the safety of these powerful generators remains entirely unexplored. Existing diffusion unlearning methods are inherently incompatible with these one-step models, as they rely on a multi-step iterative denoising process. In this work, we propose UOT-Unlearn, a novel plug-and-play class unlearning framework for one-step generative models based on the Unbalanced Optimal Transport (UOT). Our method formulates unlearning as a principled trade-off between a forget cost, which suppresses the target class, and an $f$-divergence penalty, which preserves overall generation fidelity via relaxed marginal constraints. By leveraging UOT, our method enables the probability mass of the forgotten class to be smoothly redistributed to the remaining classes, rather than collapsing into low-quality or noise-like samples. Experimental results on CIFAR-10 and ImageNet-256 demonstrate that our framework achieves superior unlearning success (PUL) and retention quality (u-FID), significantly outperforming baselines.

cs.CV↗

arXiv2026

Maximal extension on converse monogamy of entanglement for tripartite pure states

Unlike classical correlations, entanglement cannot be freely shared among multiple parties. This unique feature of quantum systems is known as the monogamy of entanglement. While it holds for all multipartite pure states, its converse -- weak entanglement between two parties enforces strong entanglement with a third party -- occurs only under specific conditions. In particular, Hayashi and Chen [Phys. Rev. A \textbf{84}, 012325 (2011)] demonstrated a qualitative version of the converse monogamy of entanglement (CMoE) for tripartite pure states by employing a hierarchy of bipartite entanglement defined through the relations among various separability criteria, and Singh and Datta [IEEE Trans. Inf. Theory \textbf{69}, 6564 (2023)] later extended this notion of the CMoE from the viewpoint of distillability under one-way or two-way classical communication. In this work, we extend their results to the CMoE with broader conditions, and furthermore show that our extensions are maximal with respect to the hierarchies they considered.

quant-ph↗