arXiv subjects

Jing Liu

Publications and source records attributed to Jing Liu.

At least 487 records · Page 27Linked to original sources

arXiv2022

Dense Text Retrieval based on Pretrained Language Models: A Survey

Text retrieval is a long-standing research topic on information seeking, where a system is required to return relevant information resources to user's queries in natural language. From classic retrieval methods to learning-based ranking functions, the underlying retrieval models have been continually evolved with the ever-lasting technical innovation. To design effective retrieval models, a key point lies in how to learn the text representation and model the relevance matching. The recent success of pretrained language models (PLMs) sheds light on developing more capable text retrieval approaches by leveraging the excellent modeling capacity of PLMs. With powerful PLMs, we can effectively learn the representations of queries and texts in the latent representation space, and further construct the semantic matching function between the dense vectors for relevance modeling. Such a retrieval approach is referred to as dense retrieval, since it employs dense vectors (a.k.a., embeddings) to represent the texts. Considering the rapid progress on dense retrieval, in this survey, we systematically review the recent advances on PLM-based dense retrieval. Different from previous surveys on dense retrieval, we take a new perspective to organize the related work by four major aspects, including architecture, training, indexing and integration, and summarize the mainstream techniques for each aspect. We thoroughly survey the literature, and include 300+ related reference papers on dense retrieval. To support our survey, we create a website for providing useful resources, and release a code repertory and toolkit for implementing dense retrieval models. This survey aims to provide a comprehensive, practical reference focused on the major progress for dense text retrieval.

cs.IR↗

arXiv2022

Transport evidence of superlattice Dirac cones in graphene monolayer on twisted boron nitride substrate

Strong band engineering in two-dimensional (2D) materials can be achieved by introducing moiré superlattices, leading to the emergence of various novel quantum phases with promising potential for future applications. Presented works to create moiré patterns have been focused on a twist embedded inside channel materials or between channel and substrate. However, the effects of a twist inside the substrate materials on the unaligned channel materials are much less explored. In this work, we report the realization of superlattice multi-Dirac cones with the coexistence of the main Dirac cone in a monolayer graphene (MLG) on a ~0.14° twisted double-layer boron nitride (tBN) substrate. Transport measurements reveal the emergence of three pairs of superlattice Dirac points around the pristine Dirac cone, featuring multiple metallic or insulating states surrounding the charge neutrality point (CNP). Displacement field tunable and electron-hole asymmetric Fermi velocities are indicated from temperature dependent measurements, along with the gapless dispersion of superlattice Dirac cones. The experimental observation of multiple Dirac cones in MLG/tBN heterostructure is supported by band structure calculations employing periodic moiré potential. Our results unveil the potential of using twisted substrate as a universal band engineering technique for 2D materials regardless of lattice matching and crystal orientations, which might pave the way for a new branch of twistronics.

cond-mat.mes-hall↗

arXiv2022

DuReader_retrieval: A Large-scale Chinese Benchmark for Passage Retrieval from Web Search Engine

In this paper, we present DuReader_retrieval, a large-scale Chinese dataset for passage retrieval. DuReader_retrieval contains more than 90K queries and over 8M unique passages from a commercial search engine. To alleviate the shortcomings of other datasets and ensure the quality of our benchmark, we (1) reduce the false negatives in development and test sets by manually annotating results pooled from multiple retrievers, and (2) remove the training queries that are semantically similar to the development and testing queries. Additionally, we provide two out-of-domain testing sets for cross-domain evaluation, as well as a set of human translated queries for for cross-lingual retrieval evaluation. The experiments demonstrate that DuReader_retrieval is challenging and a number of problems remain unsolved, such as the salient phrase mismatch and the syntactic mismatch between queries and paragraphs. These experiments also show that dense retrievers do not generalize well across domains, and cross-lingual retrieval is essentially challenging. DuReader_retrieval is publicly available at https://github.com/baidu/DuReader/tree/master/DuReader-Retrieval.

cs.CL↗

arXiv2022

AgileAvatar: Stylized 3D Avatar Creation via Cascaded Domain Bridging

Stylized 3D avatars have become increasingly prominent in our modern life. Creating these avatars manually usually involves laborious selection and adjustment of continuous and discrete parameters and is time-consuming for average users. Self-supervised approaches to automatically create 3D avatars from user selfies promise high quality with little annotation cost but fall short in application to stylized avatars due to a large style domain gap. We propose a novel self-supervised learning framework to create high-quality stylized 3D avatars with a mix of continuous and discrete parameters. Our cascaded domain bridging framework first leverages a modified portrait stylization approach to translate input selfies into stylized avatar renderings as the targets for desired 3D avatars. Next, we find the best parameters of the avatars to match the stylized avatar renderings through a differentiable imitator we train to mimic the avatar graphics engine. To ensure we can effectively optimize the discrete parameters, we adopt a cascaded relaxation-and-search pipeline. We use a human preference study to evaluate how well our method preserves user identity compared to previous work as well as manual creation. Our results achieve much higher preference scores than previous work and close to those of manual creation. We also provide an ablation study to justify the design choices in our pipeline.

cs.CV↗

arXiv2022

Quantitative study of the hardening in the Alpha Magnetic Spectrometer nuclei spectra at a few hundred GV

The most significant feature in the cosmic-ray (CR) nuclei spectra is the spectral hardening at a few hundred GV. It is important to know whether the hardening of different nuclei species is the same or not for constructing CR sources and propagation models. In this work, we collect the recently released AMS-02 CR nuclei spectra of primary species (proton, helium, carbon, oxygen, neon, magnesium, silicon, and iron), secondary species (lithium, beryllium, boron, and fluorine), and hybrid species (nitrogen, sodium, and aluminum) and study the break positions and the spectral index differences (less and greater than the break rigidity) of the spectral hardening quantitatively. The results show us that the CR nuclei spectral hardening at a few hundred GV has hybrid origins. In detail, the dominating factors of the spectral hardening for primary and secondary CR nuclei species are different: the former comes from the superposition of different kinds of CR sources, while the latter comes from the propagation process. Both of these factors influence all kinds of CR nuclei spectra, just with different weights.

astro-ph.HE↗

arXiv2022

Efficient and Accurate Quantized Image Super-Resolution on Mobile NPUs, Mobile AI & AIM 2022 challenge: Report

Image super-resolution is a common task on mobile and IoT devices, where one often needs to upscale and enhance low-resolution images and video frames. While numerous solutions have been proposed for this problem in the past, they are usually not compatible with low-power mobile NPUs having many computational and memory constraints. In this Mobile AI challenge, we address this problem and propose the participants to design an efficient quantized image super-resolution solution that can demonstrate a real-time performance on mobile NPUs. The participants were provided with the DIV2K dataset and trained INT8 models to do a high-quality 3X image upscaling. The runtime of all models was evaluated on the Synaptics VS680 Smart Home board with a dedicated edge NPU capable of accelerating quantized neural networks. All proposed solutions are fully compatible with the above NPU, demonstrating an up to 60 FPS rate when reconstructing Full HD resolution images. A detailed description of all models developed in the challenge is provided in this paper.

eess.IV↗

arXiv2022

A Multi-Transformation Evolutionary Framework for Influence Maximization in Social Networks

Influence maximization is a crucial issue for mining the deep information of social networks, which aims to select a seed set from the network to maximize the number of influenced nodes. To evaluate the influence spread of a seed set efficiently, existing studies have proposed transformations with lower computational costs to replace the expensive Monte Carlo simulation process. These alternate transformations, based on network prior knowledge, induce different search behaviors with similar characteristics to various perspectives. Specifically, it is difficult for users to determine a suitable transformation a priori. This article proposes a multi-transformation evolutionary framework for influence maximization (MTEFIM) with convergence guarantees to exploit the potential similarities and unique advantages of alternate transformations and to avoid users manually determining the most suitable one. In MTEFIM, multiple transformations are optimized simultaneously as multiple tasks. Each transformation is assigned an evolutionary solver. Three major components of MTEFIM are conducted via: 1) estimating the potential relationship across transformations based on the degree of overlap across individuals of different populations, 2) transferring individuals across populations adaptively according to the inter-transformation relationship, and 3) selecting the final output seed set containing all the transformation's knowledge. The effectiveness of MTEFIM is validated on both benchmarks and real-world social networks. The experimental results show that MTEFIM can efficiently utilize the potentially transferable knowledge across multiple transformations to achieve highly competitive performance compared to several popular IM-specific methods. The implementation of MTEFIM can be accessed at https://github.com/xiaofangxd/MTEFIM.

cs.NE↗

arXiv2022

QuanEstimation: An open-source toolkit for quantum parameter estimation

Quantum parameter estimation promises a high-precision measurement in theory, however, how to design the optimal scheme in a specific scenario, especially under a practical condition, is still a serious problem that needs to be solved case by case due to the existence of multiple mathematical bounds and optimization methods. Depending on the scenario considered, different bounds may be more or less suitable, both in terms of computational complexity and the tightness of the bound itself. At the same time, the metrological schemes provided by different optimization methods need to be tested against realization complexity, robustness, etc. Hence, a comprehensive toolkit containing various bounds and optimization methods is essential for the scheme design in quantum metrology. To fill this vacancy, here we present a Python-Julia-based open-source toolkit for quantum parameter estimation, which includes many well-used mathematical bounds and optimization methods. Utilizing this toolkit, all procedures in the scheme design, such as the optimizations of the probe state, control and measurement, can be readily and efficiently performed.

quant-ph↗

arXiv2022

Quantum Magnetometer with Dual-Coupling Optomechanics

An experimentally feasible magnetometer based on a dual-coupling optomechanical system is proposed, where the radiation-pressure coupling transduces the magnetic signal to the optical phase, and the quadratic optomechanical interaction induces a periodic squeezing effect. The latter not only amplifies the signal to be measured, but also accelerates the signal transducing rate characterized by an experimentally observable phase accumulation efficiency. In the vicinity of opto-mechanical decoupled time, the ultimate bound to the estimability of magnetic signal is proportional to $\exp(-6r)$, and then the optimized accuracy of estimation can be enhanced nearly 3 orders with a controllable squeezing parameter $r<1$. Moreover, our proposal is robust against the mechanical thermal noise, and the sensitivity of a specific measurement can reach to the order of $10^{-17}{\rm T/\sqrt{Hz}}$ in the presence of dissipations and without ground state cooling of mechanical oscillator. Our proposal fundamentally broadens the fields of quantum metrology and cavity optomechanics, with potential application for on-chip magnetic detection with high precision.

quant-ph↗

arXiv2022

Galaxy-galaxy lensing in the VOICE deep survey

The multi-band photometry of the VOICE imaging data, overlapping with 4.9 deg$^2$ of the Chandra Deep Field South (CDFS) area, enables both shape measurement and photometric redshift estimation to be the two essential quantities for weak lensing analysis. The depth of $mag_{AB}$ is up to 26.1 (5$σ$ limiting) in $r$-band. We estimate the Excess Surface Density (ESD; $ΔΣ$) based on galaxy-galaxy measurements around galaxies at lower redshift (0.10<$z_l$<0.35) while we select the background sources to be at higher redshift ranging from 0.3 to 1.5. The foreground galaxies are divided into two major categories according to their colour (blue/red), each of which has been further divided into high/low stellar mass bins. Then the halo masses of the samples are estimated by modelling the signals, and the posterior of the parameters are samples via Mote Carlo Markov Chain (MCMC) process. We compare our results with the existing Stellar-to-Halo Mass Relation (SHMR) and find that the blue low stellar mass bin (median $M_*=10^{8.31}M_\odot$) deviates from the SHMR relation whereas all other three samples agrees well with empirical curves. We interpret this discrepancy as the effect of a low star formation efficiency of the low-mass blue dwarf galaxy population dominated in the VOICE-CDFS area.

astro-ph.CO↗

arXiv2022

NTIRE 2021 Challenge on Quality Enhancement of Compressed Video: Methods and Results

This paper reviews the first NTIRE challenge on quality enhancement of compressed video, with a focus on the proposed methods and results. In this challenge, the new Large-scale Diverse Video (LDV) dataset is employed. The challenge has three tracks. Tracks 1 and 2 aim at enhancing the videos compressed by HEVC at a fixed QP, while Track 3 is designed for enhancing the videos compressed by x265 at a fixed bit-rate. Besides, the quality enhancement of Tracks 1 and 3 targets at improving the fidelity (PSNR), and Track 2 targets at enhancing the perceptual quality. The three tracks totally attract 482 registrations. In the test phase, 12 teams, 8 teams and 11 teams submitted the final results of Tracks 1, 2 and 3, respectively. The proposed methods and solutions gauge the state-of-the-art of video quality enhancement. The homepage of the challenge: https://github.com/RenYang-home/NTIRE21_VEnh

eess.IV↗

arXiv2022

Constraining first-order phase transitions with curvature perturbations

We investigate the curvature perturbations induced by the randomness of the quantum tunneling process during cosmological first-order phase transitions (PTs) and for the first time ultilize curvature perturbations to constrain the PT parameters. We find that the observations of the cosmic microwave background spectrum distortion and the ultracompact minihalo abundance can give strict constraints on the PTs below 100GeV, especially for the low-scale PTs and the weak PTs. The current constraint on the PT parameters is largely extended by the results in this work.

astro-ph.CO↗

arXiv2022

Mesa: A Memory-saving Training Framework for Transformers

There has been an explosion of interest in designing high-performance Transformers. While Transformers have delivered significant performance improvements, training such networks is extremely memory intensive owing to storing all intermediate activations that are needed for gradient computation during backpropagation, especially for long sequences. To this end, we present Mesa, a memory-saving training framework for Transformers. Specifically, Mesa uses exact activations during forward pass while storing a low-precision version of activations to reduce memory consumption during training. The low-precision activations are then dequantized during back-propagation to compute gradients. Besides, to address the heterogeneous activation distributions in the multi-head self-attention layers, we propose a head-wise activation quantization strategy, which quantizes activations based on the statistics of each head to minimize the approximation error. To further boost training efficiency, we learn quantization parameters by running estimates. More importantly, by re-investing the saved memory in employing a larger batch size or scaling up model size, we may further improve the performance under constrained computational resources. Extensive experiments on ImageNet, CIFAR-100 and ADE20K demonstrate that Mesa can achieve flexible memory-savings (up to 50%) during training while achieving comparable or even better performance. Code is available at https://github.com/ziplab/Mesa.

cs.CV↗

arXiv2022

FocusFormer: Focusing on What We Need via Architecture Sampler

Vision Transformers (ViTs) have underpinned the recent breakthroughs in computer vision. However, designing the architectures of ViTs is laborious and heavily relies on expert knowledge. To automate the design process and incorporate deployment flexibility, one-shot neural architecture search decouples the supernet training and architecture specialization for diverse deployment scenarios. To cope with an enormous number of sub-networks in the supernet, existing methods treat all architectures equally important and randomly sample some of them in each update step during training. During architecture search, these methods focus on finding architectures on the Pareto frontier of performance and resource consumption, which forms a gap between training and deployment. In this paper, we devise a simple yet effective method, called FocusFormer, to bridge such a gap. To this end, we propose to learn an architecture sampler to assign higher sampling probabilities to those architectures on the Pareto frontier under different resource constraints during supernet training, making them sufficiently optimized and hence improving their performance. During specialization, we can directly use the well-trained architecture sampler to obtain accurate architectures satisfying the given resource constraint, which significantly improves the search efficiency. Extensive experiments on CIFAR-100 and ImageNet show that our FocusFormer is able to improve the performance of the searched architectures while significantly reducing the search cost. For example, on ImageNet, our FocusFormer-Ti with 1.4G FLOPs outperforms AutoFormer-Ti by 0.5% in terms of the Top-1 accuracy.

cs.CV↗

arXiv2022

Provably Tightest Linear Approximation for Robustness Verification of Sigmoid-like Neural Networks

The robustness of deep neural networks is crucial to modern AI-enabled systems and should be formally verified. Sigmoid-like neural networks have been adopted in a wide range of applications. Due to their non-linearity, Sigmoid-like activation functions are usually over-approximated for efficient verification, which inevitably introduces imprecision. Considerable efforts have been devoted to finding the so-called tighter approximations to obtain more precise verification results. However, existing tightness definitions are heuristic and lack theoretical foundations. We conduct a thorough empirical analysis of existing neuron-wise characterizations of tightness and reveal that they are superior only on specific neural networks. We then introduce the notion of network-wise tightness as a unified tightness definition and show that computing network-wise tightness is a complex non-convex optimization problem. We bypass the complexity from different perspectives via two efficient, provably tightest approximations. The results demonstrate the promising performance achievement of our approaches over state of the art: (i) achieving up to 251.28% improvement to certified lower robustness bounds; and (ii) exhibiting notably more precise verification results on convolutional networks.

cs.LG↗

arXiv2022

Direct measurement of quantum Fisher information

In the adiabatic perturbation theory, Berry curvature is related to the generalized force, and the quantum metric tensor is linked with energy fluctuation. While the former is tested with numerous numerical results and experimental realizations, the latter is less considered. Quantum Fisher information, key to quantum precision measurement, is four times quantum metric tensor. It is difficult to relate the quantum Fisher information with some physical observable. One interesting candidate is square of the symmetric logarithmic derivative, which is usually tough to obtain both theoretically and experimentally. The adiabatic perturbation theory enlightens us to measure the energy fluctuation to directly extract the quantum Fisher information. In this article, we first adopt an alternative way to derive the link of energy fluctuation to the quantum Fisher information. Then we numerically testify the direct extraction of the quantum Fisher information based on adiabatic perturbation in two-level systems and simulate the experimental realization in nitrogen-vacancy center with experimentally practical parameters. Statistical models such as transverse field Ising model and Heisenberg spin chains are also discussed to compare with the analytical result and show the level crossing respectively. Our discussion will provide a new practical scheme to measure the quantum Fisher information, and will also benefit the quantum precision measurement and the understand of the quantum Fisher information.

quant-ph↗

arXiv2022

An Interpretability Evaluation Benchmark for Pre-trained Language Models

While pre-trained language models (LMs) have brought great improvements in many NLP tasks, there is increasing attention to explore capabilities of LMs and interpret their predictions. However, existing works usually focus only on a certain capability with some downstream tasks. There is a lack of datasets for directly evaluating the masked word prediction performance and the interpretability of pre-trained LMs. To fill in the gap, we propose a novel evaluation benchmark providing with both English and Chinese annotated data. It tests LMs abilities in multiple dimensions, i.e., grammar, semantics, knowledge, reasoning and computation. In addition, it provides carefully annotated token-level rationales that satisfy sufficiency and compactness. It contains perturbed instances for each original instance, so as to use the rationale consistency under perturbations as the metric for faithfulness, a perspective of interpretability. We conduct experiments on several widely-used pre-trained LMs. The results show that they perform very poorly on the dimensions of knowledge and computation. And their plausibility in all dimensions is far from satisfactory, especially when the rationale is short. In addition, the pre-trained LMs we evaluated are not robust on syntax-aware data. We will release this evaluation benchmark at \url{http://xyz}, and hope it can facilitate the research progress of pre-trained LMs.

cs.CL↗

arXiv2022

Learning Appearance-motion Normality for Video Anomaly Detection

Video anomaly detection is a challenging task in the computer vision community. Most single task-based methods do not consider the independence of unique spatial and temporal patterns, while two-stream structures lack the exploration of the correlations. In this paper, we propose spatial-temporal memories augmented two-stream auto-encoder framework, which learns the appearance normality and motion normality independently and explores the correlations via adversarial learning. Specifically, we first design two proxy tasks to train the two-stream structure to extract appearance and motion features in isolation. Then, the prototypical features are recorded in the corresponding spatial and temporal memory pools. Finally, the encoding-decoding network performs adversarial learning with the discriminator to explore the correlations between spatial and temporal patterns. Experimental results show that our framework outperforms the state-of-the-art methods, achieving AUCs of 98.1% and 89.8% on UCSD Ped2 and CUHK Avenue datasets.

cs.CV↗