arXiv subjects

Yilun Xu

Publications and source records attributed to Yilun Xu.

At least 37 records · Page 2Linked to original sources

arXiv2025

Superconducting qubit readout enhanced by path signature

Quantum non-demolition measurement plays an essential role in quantum technology, crucial for quantum error correction, metrology, and sensing. Conventionally, the qubit state is classified from the raw or integrated time-domain measurement record. Here, we demonstrate a method to enhance the assignment fidelity of the readout by considering the "path signature" of this measurement record, where the path signature is a mathematical tool for analyzing stochastic time series. We evaluate this approach across five different hardware setups, including those with and without readout multiplexing and parametric amplifiers, and demonstrate a significant improvement in assignment fidelity across all setups. Moreover, we show that the path signature of the measurement record provides an expressive feature set that can be used to detect and classify state transitions that occurred during the measurement, improving the prediction of the qubit state at the end of the measurement. This method has the potential to become a foundational tool for quantum technology.

quant-ph↗

arXiv2025

Think While You Generate: Discrete Diffusion with Planned Denoising

Discrete diffusion has achieved state-of-the-art performance, outperforming or approaching autoregressive models on standard benchmarks. In this work, we introduce Discrete Diffusion with Planned Denoising (DDPD), a novel framework that separates the generation process into two models: a planner and a denoiser. At inference time, the planner selects which positions to denoise next by identifying the most corrupted positions in need of denoising, including both initially corrupted and those requiring additional refinement. This plan-and-denoise approach enables more efficient reconstruction during generation by iteratively identifying and denoising corruptions in the optimal order. DDPD outperforms traditional denoiser-only mask diffusion methods, achieving superior results on language modeling benchmarks such as text8, OpenWebText, and token-based image generation on ImageNet $256 \times 256$. Notably, in language modeling, DDPD significantly reduces the performance gap between diffusion-based and autoregressive methods in terms of generative perplexity. Code is available at https://github.com/liusulin/DDPD.

cs.LG↗

arXiv2025

CopyQNN: Quantum Neural Network Extraction Attack under Varying Quantum Noise

Quantum Neural Networks (QNNs) have shown significant value across domains, with well-trained QNNs representing critical intellectual property often deployed via cloud-based QNN-as-a-Service (QNNaaS) platforms. Recent work has examined QNN model extraction attacks using classical and emerging quantum strategies. These attacks involve adversaries querying QNNaaS platforms to obtain labeled data for training local substitute QNNs that replicate the functionality of cloud-based models. However, existing approaches have largely overlooked the impact of varying quantum noise inherent in noisy intermediate-scale quantum (NISQ) computers, limiting their effectiveness in real-world settings. To address this limitation, we propose the CopyQNN framework, which employs a three-step data cleaning method to eliminate noisy data based on its noise sensitivity. This is followed by the integration of contrastive and transfer learning within the quantum domain, enabling efficient training of substitute QNNs using a limited but cleaned set of queried data. Experimental results on NISQ computers demonstrate that a practical implementation of CopyQNN significantly outperforms state-of-the-art QNN extraction attacks, achieving an average performance improvement of 8.73% across all tasks while reducing the number of required queries by 90x, with only a modest increase in hardware overhead.

quant-ph↗

arXiv2025

One-step Diffusion Models with $f$-Divergence Distribution Matching

Sampling from diffusion models involves a slow iterative process that hinders their practical deployment, especially for interactive applications. To accelerate generation speed, recent approaches distill a multi-step diffusion model into a single-step student generator via variational score distillation, which matches the distribution of samples generated by the student to the teacher's distribution. However, these approaches use the reverse Kullback-Leibler (KL) divergence for distribution matching which is known to be mode seeking. In this paper, we generalize the distribution matching approach using a novel $f$-divergence minimization framework, termed $f$-distill, that covers different divergences with different trade-offs in terms of mode coverage and training variance. We derive the gradient of the $f$-divergence between the teacher and student distributions and show that it is expressed as the product of their score differences and a weighting function determined by their density ratio. This weighting function naturally emphasizes samples with higher density in the teacher distribution, when using a less mode-seeking divergence. We observe that the popular variational score distillation approach using the reverse-KL divergence is a special case within our framework. Empirically, we demonstrate that alternative $f$-divergences, such as forward-KL and Jensen-Shannon divergences, outperform the current best variational score distillation methods across image generation tasks. In particular, when using Jensen-Shannon divergence, $f$-distill achieves current state-of-the-art one-step generation performance on ImageNet64 and zero-shot text-to-image generation on MS-COCO. Project page: https://research.nvidia.com/labs/genair/f-distill

cs.LG↗

arXiv2025

Energy-Based Diffusion Language Models for Text Generation

Despite remarkable progress in autoregressive language models, alternative generative paradigms beyond left-to-right generation are still being actively explored. Discrete diffusion models, with the capacity for parallel generation, have recently emerged as a promising alternative. Unfortunately, these models still underperform the autoregressive counterparts, with the performance gap increasing when reducing the number of sampling steps. Our analysis reveals that this degradation is a consequence of an imperfect approximation used by diffusion models. In this work, we propose Energy-based Diffusion Language Model (EDLM), an energy-based model operating at the full sequence level for each diffusion step, introduced to improve the underlying approximation used by diffusion models. More specifically, we introduce an EBM in a residual form, and show that its parameters can be obtained by leveraging a pretrained autoregressive model or by finetuning a bidirectional transformer via noise contrastive estimation. We also propose an efficient generation algorithm via parallel important sampling. Comprehensive experiments on language modeling benchmarks show that our model can consistently outperform state-of-the-art diffusion models by a significant margin, and approaches autoregressive models' perplexity. We further show that, without any generation performance drop, our framework offers a 1.3$\times$ sampling speedup over existing diffusion models. Reproduced code is available at https://github.com/MinkaiXu/Energy-Diffusion-LLM.

cs.CL↗

arXiv2025

Neural Augmentation Based Panoramic High Dynamic Range Stitching

Due to saturated regions of inputting low dynamic range (LDR) images and large intensity changes among the LDR images caused by different exposures, it is challenging to produce an information enriched panoramic LDR image without visual artifacts for a high dynamic range (HDR) scene through stitching multiple geometrically synchronized LDR images with different exposures and pairwise overlapping fields of views (OFOVs). Fortunately, the stitching of such images is innately a perfect scenario for the fusion of a physics-driven approach and a data-driven approach due to their OFOVs. Based on this new insight, a novel neural augmentation based panoramic HDR stitching algorithm is proposed in this paper. The physics-driven approach is built up using the OFOVs. Different exposed images of each view are initially generated by using the physics-driven approach, are then refined by a data-driven approach, and are finally used to produce panoramic LDR images with different exposures. All the panoramic LDR images with different exposures are combined together via a multi-scale exposure fusion algorithm to produce the final panoramic LDR image. Experimental results demonstrate the proposed algorithm outperforms existing panoramic stitching algorithms.

cs.CV↗

arXiv2025

First-principle crosstalk dynamics and Hamiltonian learning via Rabi experiments

Coherent errors constitute a significant barrier to successful large-scale quantum computation. One such error mechanism is crosstalk, which violates spatial locality or the independence of operations. We present a description of crosstalk and learn the underlying parameters by executing novel simultaneous Rabi experiments and fitting the Hamiltonian to the observed data. We use this model to predict three- and four-qubit experiments and observe excellent agreement between our theoretical predictions and experimental results. Our technique enables researchers to study the dynamics of multi-qubit circuits without performing experiments, potentially facilitating the minimization of coherent gate errors via digital pulse precompilation. Additionally, this method provides whole-chip crosstalk characterization, a useful tool for guiding quantum processor design.

quant-ph↗

arXiv2025

Quasi-Probabilistic Readout Correction of Mid-Circuit Measurements for Adaptive Feedback via Measurement Randomized Compiling

Quantum measurements are a fundamental component of quantum computing. However, on modern-day quantum computers, measurements can be more error prone than quantum gates, and are susceptible to non-unital errors as well as non-local correlations due to measurement crosstalk. While readout errors can be mitigated in post-processing, it is inefficient in the number of qubits due to a combinatorially-large number of possible states that need to be characterized. In this work, we show that measurement errors can be tailored into a simple stochastic error model using randomized compiling, enabling the efficient mitigation of readout errors via quasi-probability distributions reconstructed from the measurement of a single preparation state in an exponentially large confusion matrix. We demonstrate the scalability and power of this approach by correcting readout errors without matrix inversion on a large number of different preparation states applied to a register of eight superconducting transmon qubits. Moreover, we show that this method can be extended to mid-circuit measurements used for active feedback via quasi-probabilistic error cancellation, and demonstrate the correction of measurement errors on an ancilla qubit used to detect and actively correct bit-flip errors on an entangled memory qubit. Our approach enables the correction of readout errors on large numbers of qubits, and offers a strategy for correcting readout errors in adaptive circuits in which the results of mid-circuit measurements are used to perform conditional operations on non-local qubits in real time.

quant-ph↗

arXiv2025

Truncated Consistency Models

Consistency models have recently been introduced to accelerate sampling from diffusion models by directly predicting the solution (i.e., data) of the probability flow ODE (PF ODE) from initial noise. However, the training of consistency models requires learning to map all intermediate points along PF ODE trajectories to their corresponding endpoints. This task is much more challenging than the ultimate objective of one-step generation, which only concerns the PF ODE's noise-to-data mapping. We empirically find that this training paradigm limits the one-step generation performance of consistency models. To address this issue, we generalize consistency training to the truncated time range, which allows the model to ignore denoising tasks at earlier time steps and focus its capacity on generation. We propose a new parameterization of the consistency function and a two-stage training procedure that prevents the truncated-time training from collapsing to a trivial solution. Experiments on CIFAR-10 and ImageNet $64\times64$ datasets show that our method achieves better one-step and two-step FIDs than the state-of-the-art consistency models such as iCT-deep, using more than 2$\times$ smaller networks. Project page: https://truncated-cm.github.io/

cs.LG↗

arXiv2024

Manipulating Spectral Windings and Skin Modes through Nonconservative Couplings

The discovery of the non-Hermitian skin effect (NHSE) has revolutionized our understanding of wave propagation in non-Hermitian systems, highlighting unexpected localization effects beyond conventional theories. Here, we discover that NHSE, accompanied by multitype spectral phases, can be induced by manipulating nonconservative couplings. By characterizing the spectra through the windings of the energy bands, we demonstrate that band structures with identical, opposite, and even twisted windings can be achieved. These inequivalent types of spectra originate from the multichannel interference resulting from the interplay between conservative and nonconservative couplings. Associated with the multitype spectra, unipolar and bipolar NHSE with different eigenmode localizations can be observed. Additionally, our findings link the nonreciprocal transmission properties of the system to multiple spectral phases, indicating a connection with the skin modes. This paper paves new pathways for investigating non-Hermitian topological effects and manipulating nonreciprocal energy flow.

physics.optics↗

arXiv2024

Heavy-Tailed Diffusion Models

Diffusion models achieve state-of-the-art generation quality across many applications, but their ability to capture rare or extreme events in heavy-tailed distributions remains unclear. In this work, we show that traditional diffusion and flow-matching models with standard Gaussian priors fail to capture heavy-tailed behavior. We address this by repurposing the diffusion framework for heavy-tail estimation using multivariate Student-t distributions. We develop a tailored perturbation kernel and derive the denoising posterior based on the conditional Student-t distribution for the backward process. Inspired by $γ$-divergence for heavy-tailed distributions, we derive a training objective for heavy-tailed denoisers. The resulting framework introduces controllable tail generation using only a single scalar hyperparameter, making it easily tunable for diverse real-world distributions. As specific instantiations of our framework, we introduce t-EDM and t-Flow, extensions of existing diffusion and flow models that employ a Student-t prior. Remarkably, our approach is readily compatible with standard Gaussian diffusion models and requires only minimal code changes. Empirically, we show that our t-EDM and t-Flow outperform standard diffusion models in heavy-tail estimation on high-resolution weather datasets in which generating rare and extreme events is crucial.

cs.LG↗

arXiv2024

Hamiltonian Score Matching and Generative Flows

Classical Hamiltonian mechanics has been widely used in machine learning in the form of Hamiltonian Monte Carlo for applications with predetermined force fields. In this work, we explore the potential of deliberately designing force fields for Hamiltonian ODEs, introducing Hamiltonian velocity predictors (HVPs) as a tool for score matching and generative models. We present two innovations constructed with HVPs: Hamiltonian Score Matching (HSM), which estimates score functions by augmenting data via Hamiltonian trajectories, and Hamiltonian Generative Flows (HGFs), a novel generative model that encompasses diffusion models and flow matching as HGFs with zero force fields. We showcase the extended design space of force fields by introducing Oscillation HGFs, a generative model inspired by harmonic oscillators. Our experiments validate our theoretical insights about HSM as a novel score matching metric and demonstrate that HGFs rival leading generative modeling techniques.

cs.LG↗

arXiv2024

An Open-Source Data Storage and Visualization Platform for Collaborative Qubit Control

Developing collaborative research platforms for quantum bit control is crucial for driving innovation in the field, as they enable the exchange of ideas, data, and implementation to achieve more impactful outcomes. Furthermore, considering the high costs associated with quantum experimental setups, collaborative environments are vital for maximizing resource utilization efficiently. However, the lack of dedicated data management platforms presents a significant obstacle to progress, highlighting the necessity for essential assistive tools tailored for this purpose. Current qubit control systems are unable to handle complicated management of extensive calibration data and do not support effectively visualizing intricate quantum experiment outcomes. In this paper, we introduce QubiCSV (Qubit Control Storage and Visualization), a platform specifically designed to meet the demands of quantum computing research, focusing on the storage and analysis of calibration and characterization data in qubit control systems. As an open-source tool, QubiCSV facilitates efficient data management of quantum computing, providing data versioning capabilities for data storage and allowing researchers and programmers to interact with qubits in real time. The insightful visualization are developed to interpret complex quantum experiments and optimize qubit performance. QubiCSV not only streamlines the handling of qubit control system data but also improves the user experience with intuitive visualization features, making it a valuable asset for researchers in the quantum computing domain.

quant-ph↗

arXiv2024

Quantum phase transition in a quantum Rabi square with next-nearest-neighbor hopping

We propose a quantum Rabi square model where both the nearest-neighbor and the next-nearest-neighbor photon hopping are allowed among four quantum Rabi systems located at the vertices of a square. By tuning the next-nearest hopping strength, we realize a first-order phase transition between the antiferromagnetic superradiant phase and the frustrated superradiant phase, as well as a second-order phase transition between the normal and the superradiant phases. To understand the emergence of such phases, we show analytically that the effect induced by next-nearest hopping is equivalent to that of an artificial gauge phase. Our findings suggest that the next-nearest-neighbor hopping can serve as an alternative for the gauge phase to realize quantum control in applications of quantum simulation and quantum materials, and that our model represents a basic building block for the frustrated $J_1$-$J_2$ quantum spin model on square lattices.

quant-ph↗

arXiv2024

DisCo-Diff: Enhancing Continuous Diffusion Models with Discrete Latents

Diffusion models (DMs) have revolutionized generative learning. They utilize a diffusion process to encode data into a simple Gaussian distribution. However, encoding a complex, potentially multimodal data distribution into a single continuous Gaussian distribution arguably represents an unnecessarily challenging learning problem. We propose Discrete-Continuous Latent Variable Diffusion Models (DisCo-Diff) to simplify this task by introducing complementary discrete latent variables. We augment DMs with learnable discrete latents, inferred with an encoder, and train DM and encoder end-to-end. DisCo-Diff does not rely on pre-trained networks, making the framework universally applicable. The discrete latents significantly simplify learning the DM's complex noise-to-data mapping by reducing the curvature of the DM's generative ODE. An additional autoregressive transformer models the distribution of the discrete latents, a simple step because DisCo-Diff requires only few discrete variables with small codebooks. We validate DisCo-Diff on toy data, several image synthesis tasks as well as molecular docking, and find that introducing discrete latents consistently improves model performance. For example, DisCo-Diff achieves state-of-the-art FID scores on class-conditioned ImageNet-64/128 datasets with ODE sampler.

cs.LG↗

arXiv2024

Hardware-Efficient Randomized Compiling

Randomized compiling (RC) is an efficient method for tailoring arbitrary Markovian errors into stochastic Pauli channels. However, the standard procedure for implementing the protocol in software comes with a large experimental overhead -- namely, it scales linearly in the number of desired randomizations, each of which must be generated and measured independently. In this work, we introduce a hardware-efficient algorithm for performing RC on a cycle-by-cycle basis on the lowest level of our FPGA-based control hardware during the execution of a circuit. Importantly, this algorithm performs a different randomization per shot with zero runtime overhead beyond measuring a circuit without RC. We implement our algorithm using the QubiC control hardware, where we demonstrate significant reduction in the overall runtime of circuits implemented with RC, as well as a significantly lower variance in measured observables.

quant-ph↗

arXiv2024

Heterophilous Distribution Propagation for Graph Neural Networks

Graph Neural Networks (GNNs) have achieved remarkable success in various graph mining tasks by aggregating information from neighborhoods for representation learning. The success relies on the homophily assumption that nearby nodes exhibit similar behaviors, while it may be violated in many real-world graphs. Recently, heterophilous graph neural networks (HeterGNNs) have attracted increasing attention by modifying the neural message passing schema for heterophilous neighborhoods. However, they suffer from insufficient neighborhood partition and heterophily modeling, both of which are critical but challenging to break through. To tackle these challenges, in this paper, we propose heterophilous distribution propagation (HDP) for graph neural networks. Instead of aggregating information from all neighborhoods, HDP adaptively separates the neighbors into homophilous and heterphilous parts based on the pseudo assignments during training. The heterophilous neighborhood distribution is learned with orthogonality-oriented constraint via a trusted prototype contrastive learning paradigm. Both the homophilous and heterophilous patterns are propagated with a novel semantic-aware message passing mechanism. We conduct extensive experiments on 9 benchmark datasets with different levels of homophily. Experimental results show that our method outperforms representative baselines on heterophilous datasets.

cs.LG↗

arXiv2024

Phase transition and multistability in Dicke dimer

The exotic phase transitions and multistabilities in atom-cavity coupled systems have attracted tremendous interests recently. In this work, we investigate the effect of photon hopping between two Dicke cavities, which induces rich quantum phases for steady states and dynamic process. Starting from a generic dimer system where the two cavities are not necessarily identical, we analytically prove all possible steady-state phases, which are confirmed by numerical calculations. We then focus on the special case with two identical cavities, where all the steady states are confirmed by exact solutions. We show that photon hopping is a convenient and powerful tool to manipulate the quantum phases and induce multistable behavior in this system.

quant-ph↗