Search arXiv⌕ Search

arXiv subjects

Yu Zhang

Publications and source records attributed to Yu Zhang.

At least 991 records · Page 55Linked to original sources
arXiv2023

Engineering imaginary stark ladder in a dissipative lattice: passive $\mathcal{PT}$ symmetry, K symmetry and localized damping

We study an imaginary stark ladder model and propose a realization of the model in a dissipative chain with linearly increasing site-dependent dissipation strength. Due to the existence of a $K$-symmetry and passive $\mathcal{PT}$ symmetry, the model exhibits quite different feature from its Hermitian counterpart. With the increase of dissipation strength, the system first undergoes a passive $\mathcal{PT}$-symmetry breaking transition, with the shifted eigenvalues changing from real to complex, and then a $K$-symmetry restoring transition, characterized by the emergence of pure imaginary spectrum with equal spacing. Accordingly, the eigenstates change from $\mathcal{PT}$-unbroken extended states to the $\mathcal{PT}$-broken states, and finally to stark localized states. In the framework of the quantum open system governed by Lindblad equation with linearly increasing site-dependent dissipation, we unveil that the dynamical evolution of single particle correlation function is governed by the Hamiltonian of the imaginary stark ladder model. By studying the dynamical evolution of the density distribution under various initial states, we demonstrate that the damping dynamics displays distinct behaviors in different regions. A localized damping is observed in the strong dissipation limit.

quant-ph↗
arXiv2023

PersonaPKT: Building Personalized Dialogue Agents via Parameter-efficient Knowledge Transfer

Personalized dialogue agents (DAs) powered by large pre-trained language models (PLMs) often rely on explicit persona descriptions to maintain personality consistency. However, such descriptions may not always be available or may pose privacy concerns. To tackle this bottleneck, we introduce PersonaPKT, a lightweight transfer learning approach that can build persona-consistent dialogue models without explicit persona descriptions. By representing each persona as a continuous vector, PersonaPKT learns implicit persona-specific features directly from a small number of dialogue samples produced by the same persona, adding less than 0.1% trainable parameters for each persona on top of the PLM backbone. Empirical results demonstrate that PersonaPKT effectively builds personalized DAs with high storage efficiency, outperforming various baselines in terms of persona consistency while maintaining good response generation quality. In addition, it enhances privacy protection by avoiding explicit persona descriptions. Overall, PersonaPKT is an effective solution for creating personalized DAs that respect user privacy.

cs.CL↗
arXiv2023

Efficient Adapters for Giant Speech Models

Large pre-trained speech models are widely used as the de-facto paradigm, especially in scenarios when there is a limited amount of labeled data available. However, finetuning all parameters from the self-supervised learned model can be computationally expensive, and becomes infeasiable as the size of the model and the number of downstream tasks scales. In this paper, we propose a novel approach called Two Parallel Adapter (TPA) that is inserted into the conformer-based model pre-trained model instead. TPA is based on systematic studies of the residual adapter, a popular approach for finetuning a subset of parameters. We evaluate TPA on various public benchmarks and experiment results demonstrates its superior performance, which is close to the full finetuning on different datasets and speech tasks. These results show that TPA is an effective and efficient approach for serving large pre-trained speech models. Ablation studies show that TPA can also be pruned, especially for lower blocks.

eess.AS↗
arXiv2023

Gotta: Generative Few-shot Question Answering by Prompt-based Cloze Data Augmentation

Few-shot question answering (QA) aims at precisely discovering answers to a set of questions from context passages while only a few training samples are available. Although existing studies have made some progress and can usually achieve proper results, they suffer from understanding deep semantics for reasoning out the questions. In this paper, we develop Gotta, a Generative prOmpT-based daTa Augmentation framework to mitigate the challenge above. Inspired by the human reasoning process, we propose to integrate the cloze task to enhance few-shot QA learning. Following the recent success of prompt-tuning, we present the cloze task in the same format as the main QA task, allowing the model to learn both tasks seamlessly together to fully take advantage of the power of prompt-tuning. Extensive experiments on widely used benchmarks demonstrate that Gotta consistently outperforms competitive baselines, validating the effectiveness of our proposed prompt-tuning-based cloze task, which not only fine-tunes language models but also learns to guide reasoning in QA tasks. Further analysis shows that the prompt-based loss incorporates the auxiliary task better than the multi-task loss, highlighting the strength of prompt-tuning on the few-shot QA task.

cs.CL↗
arXiv2023

Heterformer: Transformer-based Deep Node Representation Learning on Heterogeneous Text-Rich Networks

Representation learning on networks aims to derive a meaningful vector representation for each node, thereby facilitating downstream tasks such as link prediction, node classification, and node clustering. In heterogeneous text-rich networks, this task is more challenging due to (1) presence or absence of text: Some nodes are associated with rich textual information, while others are not; (2) diversity of types: Nodes and edges of multiple types form a heterogeneous network structure. As pretrained language models (PLMs) have demonstrated their effectiveness in obtaining widely generalizable text representations, a substantial amount of effort has been made to incorporate PLMs into representation learning on text-rich networks. However, few of them can jointly consider heterogeneous structure (network) information as well as rich textual semantic information of each node effectively. In this paper, we propose Heterformer, a Heterogeneous Network-Empowered Transformer that performs contextualized text encoding and heterogeneous structure encoding in a unified model. Specifically, we inject heterogeneous structure information into each Transformer layer when encoding node texts. Meanwhile, Heterformer is capable of characterizing node/edge type heterogeneity and encoding nodes with or without texts. We conduct comprehensive experiments on three tasks (i.e., link prediction, node classification, and node clustering) on three large-scale datasets from different domains, where Heterformer outperforms competitive baselines significantly and consistently.

cs.CL↗
arXiv2023

Simple realization of a hybrid controlled-controlled-Z gate with photonic control qubits encoded via eigenstates of the photon-number parity operator

We propose a simple method to realize a hybrid controlled-controlled-Z (CCZ) gate with two photonic qubits simultaneously controlling a superconducting (SC) target qubit, by employing two microwave cavities coupled to a SC ququart (a four-level quantum system). In this proposal, each control qubit is a photonic qubit, which is encoded by two arbitrary orthogonal eigenstates (with eigenvalues 1 and -1, respectively) of the photon-number parity operator. Since the two arbitrary encoding states can take various quantum states, this proposal can be applied to realize the hybrid CCZ gate, for which the two control photonic qubits can have various encodings. The gate realization is quite simple because only a basic operation is needed. During the gate operation, the higher energy intermediate levels of the ququart are not occupied, and, thus, decoherence from these levels is greatly suppressed. We further discuss how to apply this gate to generate a hybrid Greenberger-Horne-Zeilinger (GHZ) entangled state of a SC qubit and two photonic qubits, which takes a general form. As an example, our numerical simulation demonstrates that high-fidelity generation of a cat-cat-spin hybrid GHZ state is feasible within current circuit QED technology. This proposal is quite general, which can be applied to realize the hybrid CCZ gate as well as to prepare various hybrid GHZ states of a matter qubit and two photonic qubits in other physical systems, such as two microwave or optical cavities coupled to a four-level natural or artificial atom.

quant-ph↗
arXiv2023

One-step implementation of a multi-target-qubit controlled-phase gate with photonic qubits encoded via eigenstates of the photon-number parity operator

In recent years, quantum state engineering and quantum information processing using microwave fields and photons have received increasing attention. In addition, multiqubit gates play an important role in quantum information processing. In this work, we propose to encode a photonic qubit via two arbitrary orthogonal eigenstates (with eigenvalues 1 and -1, respectively) of the photon-number parity operator. With such encoding, we then present a single-step method to realize a multi-target-qubit controlled-phase gate with one photonic qubit simultaneously controlling n-1 target photonic qubits, by employing n microwave cavities coupled to one superconducting flux qutrit. This proposal can be applied not only to implement nonhybrid multi-target-qubit controlled-phase gates using photonic qubits with various encodings, but also to realize hybrid multi-target-qubit controlled-phase gates using photonic qubits with different encodings. The gate realization requires only a single-step operation. The gate operation time does not increase with the number of target qubits. Because the qutrit remains in the ground state during the entire operation, decoherence from the qutrit is greatly suppressed. As an application, we show how to apply this gate to generate a multicavity Greenberger-Horne-Zeilinger (GHZ) entangled state with general expression. Depending on the specific encodings, we further discuss the preparation of several nonhybrid and hybrid GHZ entangled states of multiple cavities. We numerically investigate the circuit-QED experimental feasibility of creating a three-cavity spin-coherent hybrid GHZ state. This proposal can be extended to accomplish the same tasks in a wide range of physical systems, such as multiple microwave or optical cavities coupled to a three-level natural or artificial atom.

quant-ph↗
arXiv2023

Bi-LRFusion: Bi-Directional LiDAR-Radar Fusion for 3D Dynamic Object Detection

LiDAR and Radar are two complementary sensing approaches in that LiDAR specializes in capturing an object's 3D shape while Radar provides longer detection ranges as well as velocity hints. Though seemingly natural, how to efficiently combine them for improved feature representation is still unclear. The main challenge arises from that Radar data are extremely sparse and lack height information. Therefore, directly integrating Radar features into LiDAR-centric detection networks is not optimal. In this work, we introduce a bi-directional LiDAR-Radar fusion framework, termed Bi-LRFusion, to tackle the challenges and improve 3D detection for dynamic objects. Technically, Bi-LRFusion involves two steps: first, it enriches Radar's local features by learning important details from the LiDAR branch to alleviate the problems caused by the absence of height information and extreme sparsity; second, it combines LiDAR features with the enhanced Radar features in a unified bird's-eye-view representation. We conduct extensive experiments on nuScenes and ORR datasets, and show that our Bi-LRFusion achieves state-of-the-art performance for detecting dynamic objects. Notably, Radar data in these two datasets have different formats, which demonstrates the generalizability of our method. Codes are available at https://github.com/JessieW0806/BiLRFusion.

cs.CV↗
arXiv2023

Explanation Graph Generation via Generative Pre-training over Synthetic Graphs

The generation of explanation graphs is a significant task that aims to produce explanation graphs in response to user input, revealing the internal reasoning process. This task is challenging due to the significant discrepancy between unstructured user queries and structured explanation graphs. Current research commonly fine-tunes a text-based pre-trained language model on a small downstream dataset that is annotated with labeled graphs. However, due to the limited scale of available datasets, this approach may prove to be insufficient in bridging the gap between natural language text and structured graphs. In this paper, to alleviate the above limitations, we propose a novel pre-trained framework EG3P(for Explanation Graph Generation via Generative Pre-training over synthetic graphs) for the explanation graph generation task. Specifically, we first propose a text-to-graph generative task to pre-train the model with the goal of bridging the text-graph gap. Additionally, we propose an automatic corpus synthesis strategy for synthesizing a large scale of high-quality corpus, reducing the reliance on costly manual annotation methods. Experimental results on ExplaGraphs show the effectiveness of EG3P that our model surpasses all baseline systems with remarkable margins. Besides, further analysis demonstrates that EG3P is able to generate better explanation graphs on actual reasoning tasks such as CommonsenseQA and OpenbookQA.

cs.CL↗
arXiv2023

LibriTTS-R: A Restored Multi-Speaker Text-to-Speech Corpus

This paper introduces a new speech dataset called ``LibriTTS-R'' designed for text-to-speech (TTS) use. It is derived by applying speech restoration to the LibriTTS corpus, which consists of 585 hours of speech data at 24 kHz sampling rate from 2,456 speakers and the corresponding texts. The constituent samples of LibriTTS-R are identical to those of LibriTTS, with only the sound quality improved. Experimental results show that the LibriTTS-R ground-truth samples showed significantly improved sound quality compared to those in LibriTTS. In addition, neural end-to-end TTS trained with LibriTTS-R achieved speech naturalness on par with that of the ground-truth samples. The corpus is freely available for download from \url{http://www.openslr.org/141/}.

eess.AS↗
arXiv2023

Perpendicular in-plane negative magnetoresistance in ZrTe5

The unique band structure in topological materials frequently results in unusual magneto-transport phenomena, one of which is in-plane longitudinal negative magnetoresistance (NMR) with the magnetic field aligned parallel to the electrical current direction. This NMR is widely considered as a hallmark of chiral anomaly in topological materials. Here we report the observation of in-plane NMR in the topological material ZrTe5 when the in-plane magnetic field is both parallel and perpendicular to the current direction, revealing an unusual case of quantum transport beyond the chiral anomaly. We find that a general theoretical model, which considers the combined effect of Berry curvature and orbital moment, can quantitatively explain this in-plane NMR. Our results provide new insights into the understanding of in-plane NMR in topological materials.

cond-mat.mes-hall↗
arXiv2023

Embedding tensors on 3-Hom-Lie algebras

In this paper, we introduce the notion of embedding tensor on 3-Hom-Lie algebras and naturally induce 3-Hom-Leibniz algebras. Moreover, the cohomology theory of embedding tensors on 3-Hom-Lie algebras is defined. As an application, we show that if two linear deformations of an embedding tensor on a 3-Hom-Lie algebra are equivalent, then their infinitesimals belong to the same cohomology class in the first cohomology group.

math.RA↗
arXiv2023

Visually-Aware Audio Captioning With Adaptive Audio-Visual Attention

Audio captioning aims to generate text descriptions of audio clips. In the real world, many objects produce similar sounds. How to accurately recognize ambiguous sounds is a major challenge for audio captioning. In this work, inspired by inherent human multimodal perception, we propose visually-aware audio captioning, which makes use of visual information to help the description of ambiguous sounding objects. Specifically, we introduce an off-the-shelf visual encoder to extract video features and incorporate the visual features into an audio captioning system. Furthermore, to better exploit complementary audio-visual contexts, we propose an audio-visual attention mechanism that adaptively integrates audio and visual context and removes the redundant information in the latent space. Experimental results on AudioCaps, the largest audio captioning dataset, show that our proposed method achieves state-of-the-art results on machine translation metrics.

eess.AS↗
arXiv2023

Chain-of-Skills: A Configurable Model for Open-domain Question Answering

The retrieval model is an indispensable component for real-world knowledge-intensive tasks, e.g., open-domain question answering (ODQA). As separate retrieval skills are annotated for different datasets, recent work focuses on customized methods, limiting the model transferability and scalability. In this work, we propose a modular retriever where individual modules correspond to key skills that can be reused across datasets. Our approach supports flexible skill configurations based on the target domain to boost performance. To mitigate task interference, we design a novel modularization parameterization inspired by sparse Transformer. We demonstrate that our model can benefit from self-supervised pretraining on Wikipedia and fine-tuning using multiple ODQA datasets, both in a multi-task fashion. Our approach outperforms recent self-supervised retrievers in zero-shot evaluations and achieves state-of-the-art fine-tuned retrieval performance on NQ, HotpotQA and OTT-QA.

cs.CL↗
arXiv2023

Mixture-of-Expert Conformer for Streaming Multilingual ASR

End-to-end models with large capacity have significantly improved multilingual automatic speech recognition, but their computation cost poses challenges for on-device applications. We propose a streaming truly multilingual Conformer incorporating mixture-of-expert (MoE) layers that learn to only activate a subset of parameters in training and inference. The MoE layer consists of a softmax gate which chooses the best two experts among many in forward propagation. The proposed MoE layer offers efficient inference by activating a fixed number of parameters as the number of experts increases. We evaluate the proposed model on a set of 12 languages, and achieve an average 11.9% relative improvement in WER over the baseline. Compared to an adapter model using ground truth information, our MoE model achieves similar WER and activates similar number of parameters but without any language information. We further show around 3% relative WER improvement by multilingual shallow fusion.

cs.CL↗
arXiv2023

Probing active-sterile neutrino transition magnetic moments at LEP and CEPC

We consider the sterile neutrino, that is also know as heavy neutral lepton (HNL), interacting with the Standard Model (SM) active neutrino and photon via a transition magnetic moment, the so-called dipole portal, which can be induced from the more general dipole couplings which respect the full gauge symmetries of the SM. Depending on the interactions with ${\rm SU}(2)_L$ and ${\rm U(1)}_Y$ field strength tensors ${\cal W}_{μν}^a$ and $B^{μν}$, we consider four typical scenarios and probe the constraints on the couplings with photon $d_γ$ at LEP using the analyses to search monophoton signature and the measurement of $Z$ decay. We find that in the considered scenarios assuming the coupling with $Z$ boson $d_Z\neq 0$, the measurement of $Z$ decaying into photon plus invisible particles can provide stricter constraints than the monophoton searches at the LEP1. The complementary constraints to existing experiments can be provided by the LEP. We also investigate the sensitivity on the dipole portal coupling $d_γ$ from the monophoton searches at future electron colliders, such as CEPC, and find that CEPC can explore the previously unconstrained parameter space by current experiments.

hep-ph↗
arXiv2023

Patton: Language Model Pretraining on Text-Rich Networks

A real-world text corpus sometimes comprises not only text documents but also semantic links between them (e.g., academic papers in a bibliographic network are linked by citations and co-authorships). Text documents and semantic connections form a text-rich network, which empowers a wide range of downstream tasks such as classification and retrieval. However, pretraining methods for such structures are still lacking, making it difficult to build one generic model that can be adapted to various tasks on text-rich networks. Current pretraining objectives, such as masked language modeling, purely model texts and do not take inter-document structure information into consideration. To this end, we propose our PretrAining on TexT-Rich NetwOrk framework Patton. Patton includes two pretraining strategies: network-contextualized masked language modeling and masked node prediction, to capture the inherent dependency between textual attributes and network structure. We conduct experiments on four downstream tasks in five datasets from both academic and e-commerce domains, where Patton outperforms baselines significantly and consistently.

cs.CL↗