arXiv subjects

Peng Du

Publications and source records attributed to Peng Du.

At least 37 records · Page 2Linked to original sources

arXiv2025

QAMA: Scalable Quantum Annealing Multi-Head Attention Operator for Deep Learning

Attention mechanisms underpin modern deep learning, while the quadratic time and space complexity limit scalability for long sequences. To address this, Quantum Annealing Multi-Head Attention (QAMA) is proposed, a novel drop-in operator that reformulates attention as an energy-based Hamiltonian optimization problem. In this framework, token interactions are encoded into binary quadratic terms, and quantum annealing is employed to search for low-energy configurations that correspond to effective attention patterns. Unlike classical sparse or approximate attention methods that rely on hand-crafted heuristics, QAMA allows sparsity structures to emerge naturally from the optimization process. Theoretically, computational complexity is analysed through single-spin flip dynamics, providing time to solution runtime bounds that depend on the spectral properties of the annealing Hamiltonian. Empirically, evaluation on both natural language and vision benchmarks shows that, across tasks, accuracy deviates by at most 2.7 points from standard multi-head attention, while requiring only linear qubits in sequence length. Visualizations further reveal that the Hamiltonian penalty terms induce meaningful and interpretable sparsity across heads. Finally, deployment on a coherent Ising machine validates the feasibility of running QAMA on real quantum hardware, showing tangible inference-time reductions compared with classical implementations. These results highlight QAMA as a pioneering and scalable step toward integrating quantum optimization devices into deep neural architectures, providing a seamlessly integrable and hardware-compatible alternative to conventional attention mechanisms. This work has been submitted to the IEEE for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible.

quant-ph↗

arXiv2025

X-Intelligence 3.0: Training and Evaluating Reasoning LLM for Semiconductor Display

Large language models (LLMs) have recently achieved significant advances in reasoning and demonstrated their advantages in solving challenging problems. Yet, their effectiveness in the semiconductor display industry remains limited due to a lack of domain-specific training and expertise. To bridge this gap, we present X-Intelligence 3.0, the first high-performance reasoning model specifically developed for the semiconductor display industry. This model is designed to deliver expert-level understanding and reasoning for the industry's complex challenges. Leveraging a carefully curated industry knowledge base, the model undergoes supervised fine-tuning and reinforcement learning to enhance its reasoning and comprehension capabilities. To further accelerate development, we implemented an automated evaluation framework that simulates expert-level assessments. We also integrated a domain-specific retrieval-augmented generation (RAG) mechanism, resulting in notable performance gains on benchmark datasets. Despite its relatively compact size of 32 billion parameters, X-Intelligence 3.0 outperforms SOTA DeepSeek-R1-671B across multiple evaluations. This demonstrates its exceptional efficiency and establishes it as a powerful solution to the longstanding reasoning challenges faced by the semiconductor display industry.

cs.CL↗

arXiv2025

Large Scalable Cross-Domain Graph Neural Networks for Personalized Notification at LinkedIn

Notification recommendation systems are critical to driving user engagement on professional platforms like LinkedIn. Designing such systems involves integrating heterogeneous signals across domains, capturing temporal dynamics, and optimizing for multiple, often competing, objectives. Graph Neural Networks (GNNs) provide a powerful framework for modeling complex interactions in such environments. In this paper, we present a cross-domain GNN-based system deployed at LinkedIn that unifies user, content, and activity signals into a single, large-scale graph. By training on this cross-domain structure, our model significantly outperforms single-domain baselines on key tasks, including click-through rate (CTR) prediction and professional engagement. We introduce architectural innovations including temporal modeling and multi-task learning, which further enhance performance. Deployed in LinkedIn's notification system, our approach led to a 0.10% lift in weekly active users and a 0.62% improvement in CTR. We detail our graph construction process, model design, training pipeline, and both offline and online evaluations. Our work demonstrates the scalability and effectiveness of cross-domain GNNs in real-world, high-impact applications.

cs.LG↗

arXiv2025

Frequency super-resolution with quantum environment engineering in a weakly coupled nuclear-spin system

Optical super-resolution has been widely employed to beat spatial diffraction limit, which is often stated by Abbe-Rayleigh criterion. Analogously, we propose a frequency super-resolution method, which beats conventional spectral resolution limit often approximated by full width half maximum of the spectral peak, Γ. This method utilizes recently developed quantum environment engineering technique. With numerical simulations and experiments, we demonstrate the frequency super-resolution method in a three-nuclear-spin system (Trifluoroiodoethylene), by successfully decomposing a thermal state spectrum of the spin F3 into four peaks of engineered pseudo-pure states of the quantum environment. The ultimate frequency resolution reaches {\sim} 0.005 Γ. This method is potentially useful in spectral decomposition of weakly coupled nuclear spin systems and might be improved further to acquire finer frequency super-resolution by employing more advanced quantum techniques..

quant-ph↗

arXiv2025

A Continual Learning-driven Model for Accurate and Generalizable Segmentation of Clinically Comprehensive and Fine-grained Whole-body Anatomies in CT

Precision medicine in the quantitative management of chronic diseases and oncology would be greatly improved if the Computed Tomography (CT) scan of any patient could be segmented, parsed and analyzed in a precise and detailed way. However, there is no such fully annotated CT dataset with all anatomies delineated for training because of the exceptionally high manual cost, the need for specialized clinical expertise, and the time required to finish the task. To this end, we proposed a novel continual learning-driven CT model that can segment complete anatomies presented using dozens of previously partially labeled datasets, dynamically expanding its capacity to segment new ones without compromising previously learned organ knowledge. Existing multi-dataset approaches are not able to dynamically segment new anatomies without catastrophic forgetting and would encounter optimization difficulty or infeasibility when segmenting hundreds of anatomies across the whole range of body regions. Our single unified CT segmentation model, CL-Net, can highly accurately segment a clinically comprehensive set of 235 fine-grained whole-body anatomies. Composed of a universal encoder, multiple optimized and pruned decoders, CL-Net is developed using 13,952 CT scans from 20 public and 16 private high-quality partially labeled CT datasets of various vendors, different contrast phases, and pathologies. Extensive evaluation demonstrates that CL-Net consistently outperforms the upper limit of an ensemble of 36 specialist nnUNets trained per dataset with the complexity of 5% model size and significantly surpasses the segmentation accuracy of recent leading Segment Anything-style medical image foundation models by large margins. Our continual learning-driven CL-Net model would lay a solid foundation to facilitate many downstream tasks of oncology and chronic diseases using the most widely adopted CT imaging.

eess.IV↗

arXiv2025

TruePose: Human-Parsing-guided Attention Diffusion for Full-ID Preserving Pose Transfer

Pose-Guided Person Image Synthesis (PGPIS) generates images that maintain a subject's identity from a source image while adopting a specified target pose (e.g., skeleton). While diffusion-based PGPIS methods effectively preserve facial features during pose transformation, they often struggle to accurately maintain clothing details from the source image throughout the diffusion process. This limitation becomes particularly problematic when there is a substantial difference between the source and target poses, significantly impacting PGPIS applications in the fashion industry where clothing style preservation is crucial for copyright protection. Our analysis reveals that this limitation primarily stems from the conditional diffusion model's attention modules failing to adequately capture and preserve clothing patterns. To address this limitation, we propose human-parsing-guided attention diffusion, a novel approach that effectively preserves both facial and clothing appearance while generating high-quality results. We propose a human-parsing-aware Siamese network that consists of three key components: dual identical UNets (TargetNet for diffusion denoising and SourceNet for source image embedding extraction), a human-parsing-guided fusion attention (HPFA), and a CLIP-guided attention alignment (CAA). The HPFA and CAA modules can embed the face and clothes patterns into the target image generation adaptively and effectively. Extensive experiments on both the in-shop clothes retrieval benchmark and the latest in-the-wild human editing dataset demonstrate our method's significant advantages over 13 baseline approaches for preserving both facial and clothes appearance in the source image.

cs.CV↗

arXiv2024

On the Balmer spectrum of the Morel-Voevodsky category

We introduce the Morava-isotropic stable homotopy category and, more generally, the stable homotopy category of an extension $E/k$. These "local" versions of the Morel-Voevodsky stable ${\Bbb{A}}^1$-homotopy category $SH(k)$ are analogues of local motivic categories introduced in [22], but with a substantially more general notion of "isotropy". This permits to construct the, so-called, isotropic Morava points of the Balmer spectrum $\operatorname{Spc}(SH^c(k))$ of (the compact part of) the Morel-Voevodsky category. These analogues of topological Morava points are parametrized by the choice of Morava K-theory and a $K(p,m)$-equivalence class of extensions $E/k$. This provides a large supply of new points, and substantially improves our understanding of the spectrum. An interesting new feature is that the specialization among isotropic points behaves differently than in topology.

math.AG↗

arXiv2024

Enabling Efficient Batch Serving for LMaaS via Generation Length Prediction

Nowadays, large language models (LLMs) are published as a service and can be accessed by various applications via APIs, also known as language-model-as-a-service (LMaaS). Without knowing the generation length of requests, existing serving systems serve requests in a first-come, first-served (FCFS) manner with a fixed batch size, which leads to two problems that affect batch serving efficiency. First, the generation lengths of requests in a batch vary, and requests with short generation lengths must wait for requests with long generation lengths to finish during the batch serving procedure. Second, requests with longer generation lengths consume more memory during serving. Without knowing the generation lengths of batched requests, the batch size is always set small to avoid the out-of-memory (OOM) error, thus preventing the GPU from being fully utilized. In this paper, we find that a significant number of popular applications in the LMaaS scenario have a positive correlation between the generation length and the length of raw user input. Based on this observation, we propose Magnus, which can accurately predict the request generation length with the user input length, application-level, and user-level semantic features. Accordingly, Magnus can achieve high request throughput by batching requests of similar generation lengths together with adaptive batch sizes. Besides, Magnus can also schedule batches with the highest response ratio next (HRRN) policy to reduce request response time. Experiments conducted on our testbed show that Magnus improves request throughput by up to 234\% and reduces response time by up to 89.7\% compared to baselines.

cs.DC↗

arXiv2024

ZSL-RPPO: Zero-Shot Learning for Quadrupedal Locomotion in Challenging Terrains using Recurrent Proximal Policy Optimization

We present ZSL-RPPO, an improved zero-shot learning architecture that overcomes the limitations of teacher-student neural networks and enables generating robust, reliable, and versatile locomotion for quadrupedal robots in challenging terrains. We propose a new algorithm RPPO (Recurrent Proximal Policy Optimization) that directly trains recurrent neural network in partially observable environments and results in more robust training using domain randomization. Our locomotion controller supports extensive perturbation across simulation-to-reality transfer for both intrinsic and extrinsic physical parameters without further fine-tuning. This can avoid the significant decline of student's performance during simulation-to-reality transfer and therefore enhance the robustness and generalization of the locomotion controller. We deployed our controller on the Unitree A1 and Aliengo robots in real environment and exteroceptive perception is provided by either a solid-state Lidar or a depth camera. Our locomotion controller was tested in various challenging terrains like slippery surfaces, Grassy Terrain, and stairs. Our experiment results and comparison show that our approach significantly outperforms the state-of-the-art.

cs.RO↗

arXiv2024

Multi-Task Learning Using Uncertainty to Weigh Losses for Heterogeneous Face Attribute Estimation

Face images contain a wide variety of attribute information. In this paper, we propose a generalized framework for joint estimation of ordinal and nominal attributes based on information sharing. We tackle the correlation problem between heterogeneous attributes using hard parameter sharing of shallow features, and trade-off multiple loss functions by considering homoskedastic uncertainty for each attribute estimation task. This leads to optimal estimation of multiple attributes of the face and reduces the training cost of multitask learning. Experimental results on benchmarks with multiple face attributes show that the proposed approach has superior performance compared to state of the art. Finally, we discuss the bias issues arising from the proposed approach in face attribute estimation and validate its feasibility on edge systems.

cs.CV↗

arXiv2024

LiGNN: Graph Neural Networks at LinkedIn

In this paper, we present LiGNN, a deployed large-scale Graph Neural Networks (GNNs) Framework. We share our insight on developing and deployment of GNNs at large scale at LinkedIn. We present a set of algorithmic improvements to the quality of GNN representation learning including temporal graph architectures with long term losses, effective cold start solutions via graph densification, ID embeddings and multi-hop neighbor sampling. We explain how we built and sped up by 7x our large-scale training on LinkedIn graphs with adaptive sampling of neighbors, grouping and slicing of training data batches, specialized shared-memory queue and local gradient optimization. We summarize our deployment lessons and learnings gathered from A/B test experiments. The techniques presented in this work have contributed to an approximate relative improvements of 1% of Job application hearing back rate, 2% Ads CTR lift, 0.5% of Feed engaged daily active users, 0.2% session lift and 0.1% weekly active user lift from people recommendation. We believe that this work can provide practical solutions and insights for engineers who are interested in applying Graph neural networks at large scale.

cs.LG↗

arXiv2023

Reshaping limit diagrams and cofinality in higher category theory

We present some results on (co)limits of diagrams in $\infty$-categories, as well as those in $(n, 1)$-categories. In particular, we deduce a way to reshape colimit diagrams into simplicial ones, and a characterisations of $n$-cofinality for functor between $\infty$-categories. Some basics on $n$-siftedness are also briefly treated.

math.CT↗

arXiv2023

O(1) benchmarking of precise rotation in a spin-squeezed Bose-Einstein condensate

Benchmarking a high-precision quantum operation is a big challenge for many quantum systems in the presence of various noises as well as control errors. Here we propose an $O(1)$ benchmarking of a dynamically corrected rotation by taking the quantum advantage of a squeezed spin state in a spin-1 Bose-Einstein condensate. Our analytical and numerical results show that tiny rotation infidelity, defined by $1-F$ with $F$ the rotation fidelity, can be calibrated in the order of $1/N^2$ by only several measurements of the rotation error for $N$ atoms in an optimally squeezed spin state. Such an $O(1)$ benchmarking is possible not only in a spin-1 BEC but also in other many-spin or many-qubit systems if a squeezed or entangled state is available.

cond-mat.quant-gas↗

arXiv2023

Enhanced measurement precision with continuous interrogation during dynamical decoupling

Dynamical decoupling (DD) is normally ineffective when applied to DC measurement. In its straightforward implementation, DD nulls out DC signal as well while suppressing noise. This work proposes a phase relay method (PRM) that is capable of continuously interrogating the DC signal over many DD cycles. We illustrate its efficacy when applied to measurement of a weak DC magnetic field with an atomic spinor Bose-Einstein condensate. Sensitivities approaching standard quantum limit (SQL) or Heisenberg limit (HL) are potentially realizable for a coherent spin state (CSS) or a squeezed spin state (SSS) of 10,000 atoms respectively, while ambient laboratory level noise is suppressed by DD. Our work offers a practical approach to mitigate the limitations of DD to DC measurement and will like find other applications for resorting coherence in quantum sensing and quantum information processing research.

quant-ph↗

arXiv2023

Noise-resistant quantum memory enabled by Hamiltonian engineering

Nuclear spins in quantum dots are promising candidates for fast and scalable quantum memory. By utilizing the hyperfine interaction between the central electron and its surrounding nuclei, quantum information can be transferred to the collective state of the nuclei and be stored for a long time. However, nuclear spin fluctuations in a partially polarized nuclear bath deteriorate the quantum memory fidelity. Here we introduce a noise-resistant protocol to realize fast and high-fidelity quantum memory through Hamiltonian engineering. With analytics and numerics, we show that high-fidelity quantum state transfer between the electron and the nuclear spins is achievable at relatively low nuclear polarizations, due to the strong suppression of nuclear spin noises. For a realistic quantum dot with $10^4$ nuclear spins, a fidelity surpassing 80% is possible at a polarization as low as 30%. Our approach reduces the demand for high nuclear polarization, making experimentally realizing quantum memory in quantum dots more feasible.

cond-mat.mes-hall↗

arXiv2022

Optical-plug-assisted spin vortex in a $^{87}$Rb dipolar spinor Bose-Einstein condensate

Generating a spin vortex in a $^{87}$Rb dipolar spinor Bose-Einstein condensate in a controllable way is still experimentally challenging. We propose an experimentally easy and tunable way to produce spin vortex by varying the potential barrier height and the width of an additionally applied optical plug. A topological phase transition occurs from the trivial single mode approximation phase to the optical-plug-assisted-vortex one, as the barrier height increases and the width lies in an appropriate range. The optical plug causes radial density variation thus the spin vortex is favored by significantly lowering the intrinsic magnetic dipolar energy. A type of coreless spin vortex, different from the conventional polar core vortex, is predicted by our numerical results. Our proposal removes a major obstacle to investigate the topological phase transition in a $^{87}$Rb dipolar spinor BEC.

cond-mat.quant-gas↗

arXiv2022

State-selective EIT for quantum error correction in neutral atom quantum computers

We propose a way to measure the qubit state of an arbitrary sub-ensemble of atoms in an array without significantly disturbing the quantum information in the unmeasured atoms. The idea is to first site-selectively transfer atoms out of the qubit basis so that one of the two states at a time is put into an auxiliary state. Electromagnetically induced transparency (EIT) light will then protect most states while detection light is scattered from atoms in the auxiliary state, which is made immune to the EIT protection by angular momentum selection rules and carefully chosen light polarization. The two states will be measured in turn, after which it is possible to recool and return the atoms to a qubit state. These measurements can be the basis of quantum error correction.

quant-ph↗

arXiv2022

Interactive Image Synthesis with Panoptic Layout Generation

Interactive image synthesis from user-guided input is a challenging task when users wish to control the scene structure of a generated image with ease.Although remarkable progress has been made on layout-based image synthesis approaches, in order to get realistic fake image in interactive scene, existing methods require high-precision inputs, which probably need adjustment several times and are unfriendly to novice users. When placement of bounding boxes is subject to perturbation, layout-based models suffer from "missing regions" in the constructed semantic layouts and hence undesirable artifacts in the generated images. In this work, we propose Panoptic Layout Generative Adversarial Networks (PLGAN) to address this challenge. The PLGAN employs panoptic theory which distinguishes object categories between "stuff" with amorphous boundaries and "things" with well-defined shapes, such that stuff and instance layouts are constructed through separate branches and later fused into panoptic layouts. In particular, the stuff layouts can take amorphous shapes and fill up the missing regions left out by the instance layouts. We experimentally compare our PLGAN with state-of-the-art layout-based models on the COCO-Stuff, Visual Genome, and Landscape datasets. The advantages of PLGAN are not only visually demonstrated but quantitatively verified in terms of inception score, Fréchet inception distance, classification accuracy score, and coverage.

cs.CV↗