Searcharxiv⌕ Search

arXiv subjects

Xin Lu

Publications and source records attributed to Xin Lu.

At least 145 records · Page 8Linked to original sources
arXiv2023

Data-Centric Financial Large Language Models

Large language models (LLMs) show promise for natural language tasks but struggle when applied directly to complex domains like finance. LLMs have difficulty reasoning about and integrating all relevant information. We propose a data-centric approach to enable LLMs to better handle financial tasks. Our key insight is that rather than overloading the LLM with everything at once, it is more effective to preprocess and pre-understand the data. We create a financial LLM (FLLM) using multitask prompt-based finetuning to achieve data pre-processing and pre-understanding. However, labeled data is scarce for each task. To overcome manual annotation costs, we employ abductive augmentation reasoning (AAR) to automatically generate training data by modifying the pseudo labels from FLLM's own outputs. Experiments show our data-centric FLLM with AAR substantially outperforms baseline financial LLMs designed for raw text, achieving state-of-the-art on financial analysis and interpretation tasks. We also open source a new benchmark for financial analysis and interpretation. Our methodology provides a promising path to unlock LLMs' potential for complex real-world domains.

cs.CL↗
arXiv2023

Attributes Grouping and Mining Hashing for Fine-Grained Image Retrieval

In recent years, hashing methods have been popular in the large-scale media search for low storage and strong representation capabilities. To describe objects with similar overall appearance but subtle differences, more and more studies focus on hashing-based fine-grained image retrieval. Existing hashing networks usually generate both local and global features through attention guidance on the same deep activation tensor, which limits the diversity of feature representations. To handle this limitation, we substitute convolutional descriptors for attention-guided features and propose an Attributes Grouping and Mining Hashing (AGMH), which groups and embeds the category-specific visual attributes in multiple descriptors to generate a comprehensive feature representation for efficient fine-grained image retrieval. Specifically, an Attention Dispersion Loss (ADL) is designed to force the descriptors to attend to various local regions and capture diverse subtle details. Moreover, we propose a Stepwise Interactive External Attention (SIEA) to mine critical attributes in each descriptor and construct correlations between fine-grained attributes and objects. The attention mechanism is dedicated to learning discrete attributes, which will not cost additional computations in hash codes generation. Finally, the compact binary codes are learned by preserving pairwise similarities. Experimental results demonstrate that AGMH consistently yields the best performance against state-of-the-art methods on fine-grained benchmark datasets.

cs.IR↗
arXiv2023

The nature and nurture of network evolution

Although the origin of the fat-tail characteristic of the degree distribution in complex networks has been extensively researched, the underlying cause of the degree distribution characteristic across the complete range of degrees remains obscure. Here, we propose an evolution model that incorporates only two factors: the node's weight, reflecting its innate attractiveness (nature), and the node's degree, reflecting the external influences (nurture). The proposed model provides a good fit for degree distributions and degree ratio distributions of numerous real-world networks and reproduces their evolution processes. Our results indicate that the nurture factor plays a dominant role in the evolution of social networks. In contrast, the nature factor plays a dominant role in the evolution of non-social networks, suggesting that whether nodes are people determines the dominant factor influencing the evolution of real-world networks.

physics.soc-ph↗
arXiv2023

Hybrid of representation learning and reinforcement learning for dynamic and complex robotic motion planning

Motion planning is the soul of robot decision making. Classical planning algorithms like graph search and reaction-based algorithms face challenges in cases of dense and dynamic obstacles. Deep learning algorithms generate suboptimal one-step predictions that cause many collisions. Reinforcement learning algorithms generate optimal or near-optimal time-sequential predictions. However, they suffer from slow convergence, suboptimal converged results, and overfittings. This paper introduces a hybrid algorithm for robotic motion planning: long short-term memory (LSTM) pooling and skip connection for attention-based discrete soft actor critic (LSA-DSAC). First, graph network (relational graph) and attention network (attention weight) interpret the environmental state for the learning of the discrete soft actor critic algorithm. The expressive power of attention network outperforms that of graph in our task by difference analysis of these two representation methods. However, attention based DSAC faces the overfitting problem in training. Second, the skip connection method is integrated to attention based DSAC to mitigate overfitting and improve convergence speed. Third, LSTM pooling is taken to replace the sum operator of attention weigh and eliminate overfitting by slightly sacrificing convergence speed at early-stage training. Experiments show that LSA-DSAC outperforms the state-of-the-art in training and most evaluations. The physical robot is also implemented and tested in the real world.

cs.RO↗
arXiv2023

Magic momenta and three dimensional Landau levels from a three dimensional graphite moiré superlattice

Twisted bilayer graphene (TBG) and other quasi-two-dimensional moiré superlattices have attracted significant attention due to the emergence of various correlated and topological states associated with the flat bands in these systems. In this work, we theoretically explore the physical properties of a new type of \textit{three dimensional graphite moiré superlattice}, the bulk alternating twisted graphite (ATG) system with homogeneous twist angle, which is grown by in situ chemical vapor decomposition method. Compared to TBG, the bulk ATG system is bestowed with an additional wavevector degrees of freedom due to the extra dimensionality. As a result, we find that when the twist angle of bulk ATG is smaller than twice of the magic angle of TBG, there always exist ``magic momenta" at which the in-plane Fermi velocities of the moiré bands vanish. Moreover, topologically distinct flat bands of TBG at different magic angles can even co-exist at different out-of-plane wavevectors in a single bulk ATG system. Most saliently, when the twist angle is relatively large, exactly dispersionless three dimensional zeroth Landau level would emerge in the bulk ATG, which may give rise to robust three dimensional quantum Hall effects over a large range of twist angles.

cond-mat.mes-hall↗
arXiv2023

Dr.Bokeh: DiffeRentiable Occlusion-aware Bokeh Rendering

Bokeh is widely used in photography to draw attention to the subject while effectively isolating distractions in the background. Computational methods simulate bokeh effects without relying on a physical camera lens. However, in the realm of digital bokeh synthesis, the two main challenges for bokeh synthesis are color bleeding and partial occlusion at object boundaries. Our primary goal is to overcome these two major challenges using physics principles that define bokeh formation. To achieve this, we propose a novel and accurate filtering-based bokeh rendering equation and a physically-based occlusion-aware bokeh renderer, dubbed Dr.Bokeh, which addresses the aforementioned challenges during the rendering stage without the need of post-processing or data-driven approaches. Our rendering algorithm first preprocesses the input RGBD to obtain a layered scene representation. Dr.Bokeh then takes the layered representation and user-defined lens parameters to render photo-realistic lens blur. By softening non-differentiable operations, we make Dr.Bokeh differentiable such that it can be plugged into a machine-learning framework. We perform quantitative and qualitative evaluations on synthetic and real-world images to validate the effectiveness of the rendering quality and the differentiability of our method. We show Dr.Bokeh not only outperforms state-of-the-art bokeh rendering algorithms in terms of photo-realism but also improves the depth quality from depth-from-defocus.

cs.GR↗
arXiv2023

Synergistic correlated states and nontrivial topology in coupled graphene-insulator heterostructures

In this work, we study the synergistic correlated states in two distinct types of interacting electronic systems coupled by interlayer Coulomb interactions. We propose that this scenario can be realized in a type of Coulomb-coupled graphene-insulator heterostructures with gate tunable band alignment. We find that, by virtue of the interlayer Coulomb coupling between the interacting electrons in the two layers, electronic states that cannot be revealed in either individual layer would emerge in a cooperative and synergistic manner. Specifically, as a result of the band alignment, charge carriers can be transferred between graphene and the substrate under the control of gate voltages, which can yield a long-wavelength electronic crystal at the surface of the substrate. This electronic crystal exerts a superlattice Coulomb potential on the Dirac electrons in graphene, which generates subbands with reduced non-interacting Fermi velocity. As a result, $e$-$e$ Coulomb interactions within graphene would play a more important role, giving rise to a gapped Dirac state at the charge neutrality point, accompanied by interaction-enhanced Fermi velocity. Moreover, the superlattice potential can give rise to topologically nontrivial subband structures which are tunable by superlattice's constant and anisotropy. Reciprocally, the electronic crystal formed in the substrate can be substantially stabilized in such coupled bilayer heterostructure by virtue of the cooperative interlayer Coulomb coupling. We further perform high-throughput first principles calculations to identify a number of promising insulating materials as candidate substrates for graphene to demonstrate these effects.

cond-mat.mes-hall↗
arXiv2023

Text2Layer: Layered Image Generation using Latent Diffusion Model

Layer compositing is one of the most popular image editing workflows among both amateurs and professionals. Motivated by the success of diffusion models, we explore layer compositing from a layered image generation perspective. Instead of generating an image, we propose to generate background, foreground, layer mask, and the composed image simultaneously. To achieve layered image generation, we train an autoencoder that is able to reconstruct layered images and train diffusion models on the latent representation. One benefit of the proposed problem is to enable better compositing workflows in addition to the high-quality image output. Another benefit is producing higher-quality layer masks compared to masks produced by a separate step of image segmentation. Experimental results show that the proposed method is able to generate high-quality layered images and initiates a benchmark for future work.

cs.CV↗
arXiv2023

Observation of first-order quantum phase transitions and ferromagnetism in twisted double bilayer graphene

Twisted graphene multilayers are highly tunable flatband systems for developing new phases of matter. Thus far, while orbital ferromagnetism has been observed in valley polarized phases, the long-range orders of other correlated phases as well as the quantum phase transitions between different orders mostly remain unknown. Here, we report an observation of Coulomb interaction driven first-order quantum phase transitions and ferromagnetism in twisted double bilayer graphene (TDBG). At zero magnetic field, the transitions are revealed in a series of step-like abrupt resistance jumps with prominent hysteresis loop when either the displacement field (D) or the carrier density (n) is tuned across symmetry-breaking boundary near half filling, indicating a formation of ordered domains. It is worth noting that the good turnability and switching of these states gives a rise to a memory performance with a large on/off ratio. Moreover, when both spin and valley play the roles at finite magnetic field, we observe abundant first-order quantum phase transitions among normal metallic states from charge neutral point, orbital ferromagnetic states from quarter filling, and spin-polarized states from half filling. We interpret these first-order phase transitions in the picture of phase separations and spin domain percolations driven by multi-field tunable Coulomb interactions, in agreement with Lifshitz transition from Hartree-Fock calculations. The observed multi-filed tunable domain structure and its hysteresis resembles the characteristics of multiferroics, revealing intriguing magnetoelectric properties. Our result enriches the correlated phase diagram in TDBG for discovering novel exotic phases and quantum phase transitions, and it would benefit other twisted moiré systems as well.

cond-mat.mes-hall↗
arXiv2023

Don't Lose Yourself! Empathetic Response Generation via Explicit Self-Other Awareness

As a critical step to achieve human-like chatbots, empathetic response generation has attained increasing interests. Previous attempts are incomplete and not sufficient enough to elicit empathy because they only focus on the initial aspect of empathy to automatically mimic the feelings and thoughts of the user via other-awareness. However, they ignore to maintain and take the own views of the system into account, which is a crucial process to achieve the empathy called self-other awareness. To this end, we propose to generate Empathetic response with explicit Self-Other Awareness (EmpSOA). Specifically, three stages, self-other differentiation, self-other modulation and self-other generation, are devised to clearly maintain, regulate and inject the self-other aware information into the process of empathetic response generation. Both automatic and human evaluations on the benchmark dataset demonstrate the superiority of EmpSOA to generate more empathetic responses.

cs.CL↗
arXiv2023

Is ChatGPT Equipped with Emotional Dialogue Capabilities?

This report presents a study on the emotional dialogue capability of ChatGPT, an advanced language model developed by OpenAI. The study evaluates the performance of ChatGPT on emotional dialogue understanding and generation through a series of experiments on several downstream tasks. Our findings indicate that while ChatGPT's performance on emotional dialogue understanding may still lag behind that of supervised models, it exhibits promising results in generating emotional responses. Furthermore, the study suggests potential avenues for future research directions.

cs.CL↗
arXiv2023

Ground-state phase diagram of the extended two-leg $t$-$J$ ladder

Inspired by the observation of a robust $d$-wave superconducting phase driven by tuning the next-nearest-neighbor (NNN) electron hopping in recent density matrix renormalization group (DMRG) studies of six- and eight-leg $t$-$J$ model, we systematically study the phase diagram of the two-leg $t$-$J$ ladder with the NNN couplings (including NNN hopping and spin interaction) in a large region of doping level, by means of the DMRG calculations. Upon doping from half filling, we identify the Luther-Emery liquid (LEL) phase, which can be distinguished as the pairing-dominant and charge density-dominant regime by comparing the Luttinger parameter $K_ρ$. With the growing NNN couplings, pairing correlations are enhanced and correspondingly $K_ρ$ increases, driving the system from the density-dominant to the pairing-dominant regime. In the Tomonaga-Luttinger liquid (TLL) phase in the larger doping region, we identify two TLL regimes with different features of charge density correlation. At the quarter filling ($1/2$ doping level), we find that the strong dimer orders of bond energy in the open system actually decay algebraically and thus do not indicate a spontaneous translational symmetry breaking. Our results show that in the LEL phase of the two-leg ladder, the NNN couplings seem to play the similar role as that on the wider $t$-$J$ cylinder, and studies on this more accessible system can be helpful towards understanding the emergence of the remarkable $d$-wave superconducting phase on the wider system.

cond-mat.str-el↗
arXiv2023

Hybridization-Controlled Pseudogap State in the Quantum Critical Superconductor CeCoIn5

The origin of the partial suppression of the electronic density states in the enigmatic pseudogap behavior, which is at the core of understanding high-$T_c$ superconductivity, has been hotly contested as either a hallmark of preformed Cooper pairs or an incipient order of competing interactions nearby. Here, we report the quasi-particle scattering spectroscopy of the quantum critical superconductor CeCoIn$_5$, where a pseudogap with energy $Δ_g$ was manifested as a dip in the differential conductance ($dI/dV$) below the characteristic temperature of $T_g$. When subjected to external pressure, $T_g$ and $Δ_g$ gradually increase, following the trend of increase in quantum entangled hybridization between Ce 4$f$ moment and conduction electrons. On the other hand, the superconducting (SC) energy gap and its phase transition temperature shows a maximum, revealing a dome shape under pressure. The disparate dependence on pressure between the two quantum states shows that the pseudogap is less likely involved in the formation of SC Cooper pairs, but rather is controlled by Kondo hybridization, indicating that a novel type of pseudogap is realized in CeCoIn$_5$.

cond-mat.str-el↗
arXiv2022

Detecting Temporal shape changes with the Euler Characteristic Transform

Organoids are multi-cellular structures which are cultured in vitro from stem cells to resemble specific organs (e.g., brain, liver) in their three-dimensional composition. Dynamic changes in the shape and composition of these model systems can be used to understand the effect of mutations and treatments in health and disease. In this paper, we propose a new technique in the field of topological data analysis for DEtecting Temporal shape changes with the Euler Characteristic Transform (DETECT). DETECT is a rotationally invariant signature of dynamically changing shapes. We demonstrate our method on a data set of segmented videos of mouse small intestine organoid experiments and show that it outperforms classical shape descriptors. We verify our method on a synthetic organoid data set and illustrate how it generalises to 3D. We conclude that DETECT offers rigorous quantification of organoids and opens up computationally scalable methods for distinguishing different growth regimes and assessing treatment effects.

q-bio.QM↗
arXiv2022

Boosting Binary Neural Networks via Dynamic Thresholds Learning

Developing lightweight Deep Convolutional Neural Networks (DCNNs) and Vision Transformers (ViTs) has become one of the focuses in vision research since the low computational cost is essential for deploying vision models on edge devices. Recently, researchers have explored highly computational efficient Binary Neural Networks (BNNs) by binarizing weights and activations of Full-precision Neural Networks. However, the binarization process leads to an enormous accuracy gap between BNN and its full-precision version. One of the primary reasons is that the Sign function with predefined or learned static thresholds limits the representation capacity of binarized architectures since single-threshold binarization fails to utilize activation distributions. To overcome this issue, we introduce the statistics of channel information into explicit thresholds learning for the Sign Function dubbed DySign to generate various thresholds based on input distribution. Our DySign is a straightforward method to reduce information loss and boost the representative capacity of BNNs, which can be flexibly applied to both DCNNs and ViTs (i.e., DyBCNN and DyBinaryCCT) to achieve promising performance improvement. As shown in our extensive experiments. For DCNNs, DyBCNNs based on two backbones (MobileNetV1 and ResNet18) achieve 71.2% and 67.4% top1-accuracy on ImageNet dataset, outperforming baselines by a large margin (i.e., 1.8% and 1.5% respectively). For ViTs, DyBinaryCCT presents the superiority of the convolutional embedding layer in fully binarized ViTs and achieves 56.1% on the ImageNet dataset, which is nearly 9% higher than the baseline.

eess.IV↗
arXiv2022

A Unified Model for Multi-class Anomaly Detection

Despite the rapid advance of unsupervised anomaly detection, existing methods require to train separate models for different objects. In this work, we present UniAD that accomplishes anomaly detection for multiple classes with a unified framework. Under such a challenging setting, popular reconstruction networks may fall into an "identical shortcut", where both normal and anomalous samples can be well recovered, and hence fail to spot outliers. To tackle this obstacle, we make three improvements. First, we revisit the formulations of fully-connected layer, convolutional layer, as well as attention layer, and confirm the important role of query embedding (i.e., within attention layer) in preventing the network from learning the shortcut. We therefore come up with a layer-wise query decoder to help model the multi-class distribution. Second, we employ a neighbor masked attention module to further avoid the information leak from the input feature to the reconstructed output feature. Third, we propose a feature jittering strategy that urges the model to recover the correct message even with noisy inputs. We evaluate our algorithm on MVTec-AD and CIFAR-10 datasets, where we surpass the state-of-the-art alternatives by a sufficiently large margin. For example, when learning a unified model for 15 categories in MVTec-AD, we surpass the second competitor on the tasks of both anomaly detection (from 88.1% to 96.5%) and anomaly localization (from 89.5% to 96.8%). Code is available at https://github.com/zhiyuanyou/UniAD.

cs.CV↗
arXiv2022

The Equalization Losses: Gradient-Driven Training for Long-tailed Object Recognition

Long-tail distribution is widely spread in real-world applications. Due to the extremely small ratio of instances, tail categories often show inferior accuracy. In this paper, we find such performance bottleneck is mainly caused by the imbalanced gradients, which can be categorized into two parts: (1) positive part, deriving from the samples of the same category, and (2) negative part, contributed by other categories. Based on comprehensive experiments, it is also observed that the gradient ratio of accumulated positives to negatives is a good indicator to measure how balanced a category is trained. Inspired by this, we come up with a gradient-driven training mechanism to tackle the long-tail problem: re-balancing the positive/negative gradients dynamically according to current accumulative gradients, with a unified goal of achieving balance gradient ratios. Taking advantage of the simple and flexible gradient mechanism, we introduce a new family of gradient-driven loss functions, namely equalization losses. We conduct extensive experiments on a wide spectrum of visual tasks, including two-stage/single-stage long-tailed object detection (LVIS), long-tailed image classification (ImageNet-LT, Places-LT, iNaturalist), and long-tailed semantic segmentation (ADE20K). Our method consistently outperforms the baseline models, demonstrating the effectiveness and generalization ability of the proposed equalization losses. Codes will be released at https://github.com/ModelTC/United-Perception.

cs.CV↗
arXiv2022

Improving Long-tailed Object Detection with Image-Level Supervision by Multi-Task Collaborative Learning

Data in real-world object detection often exhibits the long-tailed distribution. Existing solutions tackle this problem by mitigating the competition between the head and tail categories. However, due to the scarcity of training samples, tail categories are still unable to learn discriminative representations. Bringing more data into the training may alleviate the problem, but collecting instance-level annotations is an excruciating task. In contrast, image-level annotations are easily accessible but not fully exploited. In this paper, we propose a novel framework CLIS (multi-task Collaborative Learning with Image-level Supervision), which leverage image-level supervision to enhance the detection ability in a multi-task collaborative way. Specifically, there are an object detection task (consisting of an instance-classification task and a localization task) and an image-classification task in our framework, responsible for utilizing the two types of supervision. Different tasks are trained collaboratively by three key designs: (1) task-specialized sub-networks that learn specific representations of different tasks without feature entanglement. (2) a siamese sub-network for the image-classification task that shares its knowledge with the instance-classification task, resulting in feature enrichment of detectors. (3) a contrastive learning regularization that maintains representation consistency, bridging feature gaps of different supervision. Extensive experiments are conducted on the challenging LVIS dataset. Without sophisticated loss engineering, CLIS achieves an overall AP of 31.1 with 10.1 point improvement on tail categories, establishing a new state-of-the-art. Code will be at https://github.com/waveboo/CLIS.

cs.CV↗