arXiv subjects

Tony Liu

Publications and source records attributed to Tony Liu.

8 recordsLinked to original sources

arXiv2026

Kinetic Lifshitz invariants and dynamics of nonreciprocal fluctuations in superconductors

We derive the generalized time-dependent Ginzburg-Landau theory of a disordered noncentrosymmetric superconductor from the Keldysh nonlinear sigma model, using a two-dimensional electron gas with Rashba spin-orbit coupling and an in-plane Zeeman field as a minimal model. On the thermodynamic side we construct the Lifshitz invariants of the free energy, the linear and cubic gradient terms and the momentum-odd part of the quartic vertex, and trace their dependence on disorder. These couplings are governed by a single closed-form kernel controlled by the ratio of the Dyakonov-Perel spin-relaxation rate to temperature, interpolating between the weak-relaxation regime, where the invariants are suppressed, and the relaxation-dominated regime, where the helical modulation of the order parameter saturates at a universal, disorder-independent value. This crossover reconciles conflicting results for the magnetoelectric couplings of dirty Rashba superconductors. Because the theory is formulated on the Keldysh contour, it also determines the dissipative dynamics: the relaxation rate of a fluctuation with pair momentum $\mathbf{q}$, and hence, by the fluctuation-dissipation theorem, the Langevin noise power, acquires a term odd in $\mathbf{q}$ and odd in the magnetic field. The structure of this kinetic Lifshitz invariant is dictated by Onsager reciprocity: friction and noise renormalize in lockstep, so equal-time fluctuations remain Gibbsian while the dynamics are nonreciprocal. As applications we compute the superconducting diode efficiency near $T_c$, where the cubic invariant competes with the quartic vertex and with even higher-gradient terms rendered odd by the helical shift, reversing the sign of the diode coefficient, and the fluctuation-induced magnetochiral anisotropy above $T_c$, where the current-resolved nonreciprocal resistance forms a plateau across the Gaussian regime.

cond-mat.supr-con↗

arXiv2026

When Does Disaggregation Pay? Simulating Prefill--Decode--Attention--FFN Specialization for Agentic LLM Inference

Agentic inference now dominates the LLM inference landscape, requiring LLMs to actively engage in multi-turn interactions with tool-calling capabilities. This introduces a more complex workload for the underlying inference system: serving stages such as prefill and decode exhibit substantially different behaviors and demand distinct compute and memory-bandwidth capabilities. As a result, a single homogeneous GPU system now struggles to support agentic inference, motivating an industry shift toward heterogeneous systems with disaggregated serving capabilities, such as the emerging Vera-Rubin platform with GPUs and Groq LPUs. However, the question of what the optimal hardware should look like for each component in a heterogeneous system remains underexplored. To this end, we propose a novel simulation framework for disaggregated serving, termed \textbf{HeteroPanacea}, that enables system-level simulation across three dimensions: 1) disaggregated quantization, 2) automated intra- and inter-device parallelization scheduling, and 3) PDAF (prefill-decode-attention-FFN) NPU architectural heterogeneity. By combining these three axes, we provide a cross-stack simulation framework for future heterogeneous agentic serving systems. We confirm the benefit of Prefill Decode disaggregation, simulating increased serving throughput by up to 75\% compared to traditional serving with current GPUs and demonstrate 4 way Prefill Decode Attention FFN disaggregation is the most consistent for increasing throughput across different models, assuming custom NPUs. We also investigate the relationship between model architecture and gain from disaggregation by running a set of ablation studies.

cs.DC↗

arXiv2026

Anharmonicity and Charge-Noise Sensitivity of Fraunhofer Qubit

We present a theory of a flux-tunable superconducting qubit, the "Fraunhofer qubit," based on the Fraunhofer interference in a wide ballistic Josephson junction. As magnetic flux threads the junction, the Josephson potential is effectively averaged over a phase window proportional to flux. For perfectly transmitting junctions, as flux approaches one flux quantum h/2e, the flux averaging transforms the potential near its minimum from a quadratic to a triangular shape, resulting in significantly enhanced anharmonicity. This enhancement persists for junctions with lower transparency conducting channels. Microscopic tight-binding simulations that include inhomogeneous electrostatic potential and disorder confirm the enhancement of anharmonicity. These results establish a framework for flux control in hybrid superconducting circuits, providing an operating point where anharmonicity and charge-noise protection can be optimally balanced.

cond-mat.mes-hall↗

arXiv2025

Admittance and critical current of nonreciprocal Josephson junctions

We investigate the nonequilibrium current response in diffusive superconductor-normal-metal-superconductor junctions subjected to a low-frequency AC voltage. Using a kinetic description based on the adiabatic motion of Andreev bound states, we derive a general expression for the admittance of a junction under a DC phase bias, formulated entirely in terms of the phase-dependent density of states induced by the proximity effect. A numerical solution of the full nonlinear Usadel equations that describe the dynamics of the junction is presented. The obtained results for the admittance and the Josephson current-phase relation apply to two-dimensional planar junctions with Rashba spin-orbit coupling and an in-plane Zeeman field, as well as to Josephson junctions formed with topological insulator surface states as the normal layer. The frequency dependence of the admittance captures the crossover between the hydrodynamic and collisionless regimes, distinguished by the relation between the drive frequency and the inelastic relaxation rate in the normal region.

cond-mat.supr-con↗

arXiv2023

Object Detector Differences when using Synthetic and Real Training Data

To train well-performing generalizing neural networks, sufficiently large and diverse datasets are needed. Collecting data while adhering to privacy legislation becomes increasingly difficult and annotating these large datasets is both a resource-heavy and time-consuming task. An approach to overcome these difficulties is to use synthetic data since it is inherently scalable and can be automatically annotated. However, how training on synthetic data affects the layers of a neural network is still unclear. In this paper, we train the YOLOv3 object detector on real and synthetic images from city environments. We perform a similarity analysis using Centered Kernel Alignment (CKA) to explore the effects of training on synthetic data on a layer-wise basis. The analysis captures the architecture of the detector while showing both different and similar patterns between different models. With this similarity analysis we want to give insights on how training synthetic data affects each layer and to give a better understanding of the inner workings of complex neural networks. The results show that the largest similarity between a detector trained on real data and a detector trained on synthetic data was in the early layers, and the largest difference was in the head part. The results also show that no major difference in performance or similarity could be seen between frozen and unfrozen backbone.

cs.CV↗

arXiv2023

A large language model-assisted education tool to provide feedback on open-ended responses

Open-ended questions are a favored tool among instructors for assessing student understanding and encouraging critical exploration of course material. Providing feedback for such responses is a time-consuming task that can lead to overwhelmed instructors and decreased feedback quality. Many instructors resort to simpler question formats, like multiple-choice questions, which provide immediate feedback but at the expense of personalized and insightful comments. Here, we present a tool that uses large language models (LLMs), guided by instructor-defined criteria, to automate responses to open-ended questions. Our tool delivers rapid personalized feedback, enabling students to quickly test their knowledge and identify areas for improvement. We provide open-source reference implementations both as a web application and as a Jupyter Notebook widget that can be used with instructional coding or math notebooks. With instructor guidance, LLMs hold promise to enhance student learning outcomes and elevate instructional methodologies.

cs.CY↗

arXiv2022

Negative critical currents in single-channel Josephson junctions

We argue that negative critical currents arise generically in Josephson junctions formed by single channel conductors. Specifically, we theoretically study the Josephson coupling between two superconducting leads connected by a one-dimensional conductor in the Coulomb blockade regime. We show that in the clean regime the sign of the critical current alternates with the number of electrons in the normal region. For odd occupancy the critical current is negative even when the number of electrons on the conductor is large.

cond-mat.supr-con↗

arXiv2020

AIDX: Adaptive Inference Scheme to Mitigate State-Drift in Memristive VMM Accelerators

An adaptive inference method for crossbar (AIDX) is presented based on an optimization scheme for adjusting the duration and amplitude of input voltage pulses. AIDX minimizes the long-term effects of memristance drift on artificial neural network accuracy. The sub-threshold behavior of memristor has been modeled and verified by comparing with fabricated device data. The proposed method has been evaluated by testing on different network structures and applications, e.g., image reconstruction and classification tasks. The results showed an average of 60% improvement in convolutional neural network (CNN) performance on CIFAR10 dataset after 10000 inference operations as well as 78.6% error reduction in image reconstruction.

cs.ET↗