arXiv subjects

Kai Zhang

Publications and source records attributed to Kai Zhang.

At least 379 records · Page 21Linked to original sources

arXiv2024

MagicBrush: A Manually Annotated Dataset for Instruction-Guided Image Editing

Text-guided image editing is widely needed in daily life, ranging from personal use to professional applications such as Photoshop. However, existing methods are either zero-shot or trained on an automatically synthesized dataset, which contains a high volume of noise. Thus, they still require lots of manual tuning to produce desirable outcomes in practice. To address this issue, we introduce MagicBrush (https://osu-nlp-group.github.io/MagicBrush/), the first large-scale, manually annotated dataset for instruction-guided real image editing that covers diverse scenarios: single-turn, multi-turn, mask-provided, and mask-free editing. MagicBrush comprises over 10K manually annotated triplets (source image, instruction, target image), which supports trainining large-scale text-guided image editing models. We fine-tune InstructPix2Pix on MagicBrush and show that the new model can produce much better images according to human evaluation. We further conduct extensive experiments to evaluate current image editing baselines from multiple dimensions including quantitative, qualitative, and human evaluations. The results reveal the challenging nature of our dataset and the gap between current baselines and real-world editing needs.

cs.CV↗

arXiv2024

Universal spectral moment theorem and its applications in non-Hermitian systems

The high sensitivity of the spectrum and wavefunctions to boundary conditions, termed the non-Hermitian skin effect, represents a fundamental aspect of non-Hermitian systems. While it endows non-Hermitian systems with unprecedented physical properties, it presents notable obstacles in grasping universal properties that are robust against microscopic details and boundary conditions. In this Letter, we introduce a pivotal theorem: in the thermodynamic limit, for any non-Hermitian systems with finite-range interactions, all spectral moments are invariant quantities, independent of boundary conditions, posing strong constraints on the spectrum. Utilizing this invariance, we propose a new criterion for bulk dynamical phases based on experimentally observable features and applicable to any dimensions and any boundary conditions. Based on this criterion, we define the bulk dispersive-to-proliferative phase transition, which is distinct from the real-to-complex spectral transition and contrary to traditional expectations. We verify these findings in 1D and 2D lattice models.

cond-mat.mes-hall↗

arXiv2024

Pointwise regularity for locally uniformly elliptic equations and applications

In this paper, we study the regularity for viscosity solutions of locally uniformly elliptic equations and obtain a series of interior pointwise $C^{k,α}$ ($k\geq 1$, $0<α<1$) regularity with smallness assumptions on the solution and the right-hand term. As applications, we obtain various interior pointwise regularity for several classical elliptic equations, i.e., the prescribed mean curvature equation, the Monge-Ampère equation, the $k$-Hessian equations, the $k$-Hessian quotient equations and the Lagrangian mean curvature equation. Moreover, the smallness assumptions are necessary in most cases (Remark 2.6, Remark 3.5, Remark 4.7, Remark 5.4 and Remark 6.5).

math.AP↗

arXiv2024

Point Resampling and Ray Transformation Aid to Editable NeRF Models

In NeRF-aided editing tasks, object movement presents difficulties in supervision generation due to the introduction of variability in object positions. Moreover, the removal operations of certain scene objects often lead to empty regions, presenting challenges for NeRF models in inpainting them effectively. We propose an implicit ray transformation strategy, allowing for direct manipulation of the 3D object's pose by operating on the neural-point in NeRF rays. To address the challenge of inpainting potential empty regions, we present a plug-and-play inpainting module, dubbed differentiable neural-point resampling (DNR), which interpolates those regions in 3D space at the original ray locations within the implicit space, thereby facilitating object removal & scene inpainting tasks. Importantly, employing DNR effectively narrows the gap between ground truth and predicted implicit features, potentially increasing the mutual information (MI) of the features across rays. Then, we leverage DNR and ray transformation to construct a point-based editable NeRF pipeline PR^2T-NeRF. Results primarily evaluated on 3D object removal & inpainting tasks indicate that our pipeline achieves state-of-the-art performance. In addition, our pipeline supports high-quality rendering visualization for diverse editing operations without necessitating extra supervision.

cs.CV↗

arXiv2024

Loss-induced universal one-way transport in periodically driven systems

In this paper, we show that a periodically driven Aubry-André-Harper model with imbalanced on-site gain/loss supports universal one-way transport that is immune to impurities and independent of initial excitations. We reveal the underlying mechanism that the periodic driving gives rise to the non-Hermitian skin effect in the effective Floquet Hamiltonian, thereby causing universal non-reciprocal transport. Additionally, we probe the time-average decay rate of the propagator under long-time bulk dynamics as a signature of the Floquet emergent non-Hermitian skin effect. Our results provide a feasible and controllable way to realize universal one-way transport that is easily accessible to experiments.

quant-ph↗

arXiv2024

GS-LRM: Large Reconstruction Model for 3D Gaussian Splatting

We propose GS-LRM, a scalable large reconstruction model that can predict high-quality 3D Gaussian primitives from 2-4 posed sparse images in 0.23 seconds on single A100 GPU. Our model features a very simple transformer-based architecture; we patchify input posed images, pass the concatenated multi-view image tokens through a sequence of transformer blocks, and decode final per-pixel Gaussian parameters directly from these tokens for differentiable rendering. In contrast to previous LRMs that can only reconstruct objects, by predicting per-pixel Gaussians, GS-LRM naturally handles scenes with large variations in scale and complexity. We show that our model can work on both object and scene captures by training it on Objaverse and RealEstate10K respectively. In both scenarios, the models outperform state-of-the-art baselines by a wide margin. We also demonstrate applications of our model in downstream 3D generation tasks. Our project webpage is available at: https://sai-bi.github.io/project/gs-lrm/ .

cs.CV↗

arXiv2024

Unleashing the Power of Multi-Task Learning: A Comprehensive Survey Spanning Traditional, Deep, and Pretrained Foundation Model Eras

MTL is a learning paradigm that effectively leverages both task-specific and shared information to address multiple related tasks simultaneously. In contrast to STL, MTL offers a suite of benefits that enhance both the training process and the inference efficiency. MTL's key advantages encompass streamlined model architecture, performance enhancement, and cross-domain generalizability. Over the past twenty years, MTL has become widely recognized as a flexible and effective approach in various fields, including CV, NLP, recommendation systems, disease prognosis and diagnosis, and robotics. This survey provides a comprehensive overview of the evolution of MTL, encompassing the technical aspects of cutting-edge methods from traditional approaches to deep learning and the latest trend of pretrained foundation models. Our survey methodically categorizes MTL techniques into five key areas: regularization, relationship learning, feature propagation, optimization, and pre-training. This categorization not only chronologically outlines the development of MTL but also dives into various specialized strategies within each category. Furthermore, the survey reveals how the MTL evolves from handling a fixed set of tasks to embracing a more flexible approach free from task or modality constraints. It explores the concepts of task-promptable and -agnostic training, along with the capacity for ZSL, which unleashes the untapped potential of this historically coveted learning paradigm. Overall, we hope this survey provides the research community with a comprehensive overview of the advancements in MTL from its inception in 1997 to the present in 2023. We address present challenges and look ahead to future possibilities, shedding light on the opportunities and potential avenues for MTL research in a broad manner. This project is publicly available at https://github.com/junfish/Awesome-Multitask-Learning.

cs.LG↗

arXiv2024

Coherent control of an optical tweezer phonon laser

The creation and manipulation of coherence continues to capture the attention of scientists and engineers. The optical laser is a canonical example of a system that, in principle, exhibits complete coherence. Recent research has focused on the creation of coherent, laser-like states in other physical systems. The phonon laser is one example where it is possible to amplify self-sustained mechanical oscillations. A single mode phonon laser in a levitated optical tweezer has been demonstrated through appropriate balance of active feedback gain and damping. In this work, coherent control of the dynamics of an optical tweezer phonon laser is used to share coherence between its different modes of oscillation, creating a multimode phonon laser. The coupling of the modes is achieved by periodically rotating the asymmetric optical potential in the transverse focal plane of the trapping beam via trap laser polarization rotation. The presented theory and experiment demonstrate that coherence can be transferred across different modes of an optical tweezer phonon laser, and are a step toward using these systems for precision measurement and quantum information processing.

physics.optics↗

arXiv2024

Sora: A Review on Background, Technology, Limitations, and Opportunities of Large Vision Models

Sora is a text-to-video generative AI model, released by OpenAI in February 2024. The model is trained to generate videos of realistic or imaginative scenes from text instructions and show potential in simulating the physical world. Based on public technical reports and reverse engineering, this paper presents a comprehensive review of the model's background, related technologies, applications, remaining challenges, and future directions of text-to-video AI models. We first trace Sora's development and investigate the underlying technologies used to build this "world simulator". Then, we describe in detail the applications and potential impact of Sora in multiple industries ranging from film-making and education to marketing. We discuss the main challenges and limitations that need to be addressed to widely deploy Sora, such as ensuring safe and unbiased video generation. Lastly, we discuss the future development of Sora and video generation models in general, and how advancements in the field could enable new ways of human-AI interaction, boosting productivity and creativity of video generation.

cs.CV↗

arXiv2024

Equivariant Multi-Modality Image Fusion

Multi-modality image fusion is a technique that combines information from different sensors or modalities, enabling the fused image to retain complementary features from each modality, such as functional highlights and texture details. However, effective training of such fusion models is challenging due to the scarcity of ground truth fusion data. To tackle this issue, we propose the Equivariant Multi-Modality imAge fusion (EMMA) paradigm for end-to-end self-supervised learning. Our approach is rooted in the prior knowledge that natural imaging responses are equivariant to certain transformations. Consequently, we introduce a novel training paradigm that encompasses a fusion module, a pseudo-sensing module, and an equivariant fusion module. These components enable the net training to follow the principles of the natural sensing-imaging process while satisfying the equivariant imaging prior. Extensive experiments confirm that EMMA yields high-quality fusion results for infrared-visible and medical images, concurrently facilitating downstream multi-modal segmentation and detection tasks. The code is available at https://github.com/Zhaozixiang1228/MMIF-EMMA.

cs.CV↗

arXiv2024

Edge theory of non-Hermitian skin modes in higher dimensions

In this paper, we establish an effective edge theory to characterize non-Hermitian edge-skin modes in higher dimensions. We begin by proposing a bulk projection criterion to straightforwardly identify the localized edges of skin modes. Through an exact mapping, we show that the edge-skin mode shares the same bulk-boundary correspondence and localization characteristics as the zero-energy edge states in a Hermitian semimetal under open-boundary conditions, bridging the gap between non-Hermitian edge-skin effect and Hermitian semimetals. Another key finding is the introduction of ``skewness,'' a term we proposed to describe the characteristic decay direction of skin mode from the localized edge into the bulk. Remarkably, we demonstrate that skewness is an intrinsic quantity of the skin mode and can be analytically determined using the corresponding cylinder-geometry bulk Hamiltonian, without requiring any boundary details. Furthermore, we reveal that in the edge-skin effect, the spectrum exhibits anomalous spectral sensitivity to weak local disturbances, a feature that crucially distinguishes it from the corner-skin effect.

cond-mat.mes-hall↗

arXiv2024

Reformulating Sequential Recommendation: Learning Dynamic User Interest with Content-enriched Language Modeling

Recommender systems are indispensable in the realm of online applications, and sequential recommendation has enjoyed considerable prevalence due to its capacity to encapsulate the dynamic shifts in user interests. However, previous sequential modeling methods still have limitations in capturing contextual information. The primary reason is the lack of understanding of domain-specific knowledge and item-related textual content. Fortunately, the emergence of powerful language models has unlocked the potential to incorporate extensive world knowledge into recommendation algorithms, enabling them to go beyond simple item attributes and truly understand the world surrounding user preferences. To achieve this, we propose LANCER, which leverages the semantic understanding capabilities of pre-trained language models to generate personalized recommendations. Our approach bridges the gap between language models and recommender systems, resulting in more human-like recommendations. We demonstrate the effectiveness of our approach through a series of experiments conducted on multiple benchmark datasets, showing promising results and providing valuable insights into the influence of our model on sequential recommendation tasks. Furthermore, our experimental codes are publicly available at https://github.com/Gnimixy/lancer.

cs.IR↗

arXiv2024

Objects With Lighting: A Real-World Dataset for Evaluating Reconstruction and Rendering for Object Relighting

Reconstructing an object from photos and placing it virtually in a new environment goes beyond the standard novel view synthesis task as the appearance of the object has to not only adapt to the novel viewpoint but also to the new lighting conditions and yet evaluations of inverse rendering methods rely on novel view synthesis data or simplistic synthetic datasets for quantitative analysis. This work presents a real-world dataset for measuring the reconstruction and rendering of objects for relighting. To this end, we capture the environment lighting and ground truth images of the same objects in multiple environments allowing to reconstruct the objects from images taken in one environment and quantify the quality of the rendered views for the unseen lighting environments. Further, we introduce a simple baseline composed of off-the-shelf methods and test several state-of-the-art methods on the relighting task and show that novel view synthesis is not a reliable proxy to measure performance. Code and dataset are available at https://github.com/isl-org/objects-with-lighting .

cs.CV↗

arXiv2024

Mind's Mirror: Distilling Self-Evaluation Capability and Comprehensive Thinking from Large Language Models

Large language models (LLMs) have achieved remarkable advancements in natural language processing. However, the massive scale and computational demands of these models present formidable challenges when considering their practical deployment in resource-constrained environments. While techniques such as chain-of-thought (CoT) distillation have displayed promise in distilling LLMs into small language models (SLMs), there is a risk that distilled SLMs may still inherit flawed reasoning and hallucinations from LLMs. To address these issues, we propose a twofold methodology: First, we introduce a novel method for distilling the self-evaluation capability from LLMs into SLMs, aiming to mitigate the adverse effects of flawed reasoning and hallucinations inherited from LLMs. Second, we advocate for distilling more comprehensive thinking by incorporating multiple distinct CoTs and self-evaluation outputs, to ensure a more thorough and robust knowledge transfer into SLMs. Experiments on three NLP benchmarks demonstrate that our method significantly improves the performance of distilled SLMs, offering a new perspective for developing more effective and efficient SLMs in resource-constrained environments.

cs.CL↗

arXiv2024

LLM-based Medical Assistant Personalization with Short- and Long-Term Memory Coordination

Large Language Models (LLMs), such as GPT3.5, have exhibited remarkable proficiency in comprehending and generating natural language. On the other hand, medical assistants hold the potential to offer substantial benefits for individuals. However, the exploration of LLM-based personalized medical assistant remains relatively scarce. Typically, patients converse differently based on their background and preferences which necessitates the task of enhancing user-oriented medical assistant. While one can fully train an LLM for this objective, the resource consumption is unaffordable. Prior research has explored memory-based methods to enhance the response with aware of previous mistakes for new queries during a dialogue session. We contend that a mere memory module is inadequate and fully training an LLM can be excessively costly. In this study, we propose a novel computational bionic memory mechanism, equipped with a parameter-efficient fine-tuning (PEFT) schema, to personalize medical assistants.

cs.CL↗

arXiv2024

Deep Equilibrium Diffusion Restoration with Parallel Sampling

Diffusion model-based image restoration (IR) aims to use diffusion models to recover high-quality (HQ) images from degraded images, achieving promising performance. Due to the inherent property of diffusion models, most existing methods need long serial sampling chains to restore HQ images step-by-step, resulting in expensive sampling time and high computation costs. Moreover, such long sampling chains hinder understanding the relationship between inputs and restoration results since it is hard to compute the gradients in the whole chains. In this work, we aim to rethink the diffusion model-based IR models through a different perspective, i.e., a deep equilibrium (DEQ) fixed point system, called DeqIR. Specifically, we derive an analytical solution by modeling the entire sampling chain in these IR models as a joint multivariate fixed point system. Based on the analytical solution, we can conduct parallel sampling and restore HQ images without training. Furthermore, we compute fast gradients via DEQ inversion and found that initialization optimization can boost image quality and control the generation direction. Extensive experiments on benchmarks demonstrate the effectiveness of our method on typical IR tasks and real-world settings.

cs.CV↗

arXiv2024

PURPLE: Making a Large Language Model a Better SQL Writer

Large Language Model (LLM) techniques play an increasingly important role in Natural Language to SQL (NL2SQL) translation. LLMs trained by extensive corpora have strong natural language understanding and basic SQL generation abilities without additional tuning specific to NL2SQL tasks. Existing LLMs-based NL2SQL approaches try to improve the translation by enhancing the LLMs with an emphasis on user intention understanding. However, LLMs sometimes fail to generate appropriate SQL due to their lack of knowledge in organizing complex logical operator composition. A promising method is to input the LLMs with demonstrations, which include known NL2SQL translations from various databases. LLMs can learn to organize operator compositions from the input demonstrations for the given task. In this paper, we propose PURPLE (Pre-trained models Utilized to Retrieve Prompts for Logical Enhancement), which improves accuracy by retrieving demonstrations containing the requisite logical operator composition for the NL2SQL task on hand, thereby guiding LLMs to produce better SQL translation. PURPLE achieves a new state-of-the-art performance of 80.5% exact-set match accuracy and 87.8% execution match accuracy on the validation set of the popular NL2SQL benchmark Spider. PURPLE maintains high accuracy across diverse benchmarks, budgetary constraints, and various LLMs, showing robustness and cost-effectiveness.

cs.DB↗

arXiv2024

The ALMaQUEST Survey XV: The Dependence of the Molecular-to-Atomic Gas Ratios on Resolved Optical Diagnostics

The atomic-to-molecular gas conversion is a critical step in the baryon cycle of galaxies, which sets the initial conditions for subsequent star formation and influences the multi-phase interstellar medium. We compiled a sample of 94 nearby galaxies with observations of multi-phase gas contents by utilizing public H I, CO, and optical IFU data from the MaNGA survey together with new FAST H I observations. In agreement with previous results, our sample shows that the global molecular-to-atomic gas ratio ($R_{\rm mol} \equiv$ log $M_{\rm H_2}/M_{\rm H\ I}$) is correlated with the global stellar mass surface density $μ_*$ with a Kendall's $τ$ coefficient of 0.25 and $p < 10^{-3}$, less tightly but still correlated with stellar mass and NUV$-$ r color, and not related to the specific star formation rate (sSFR). The cold gas distribution and kinematics inferred from the H I and CO global profile asymmetry and shape do not significantly rely on $R_{\rm mol}$. Thanks to the availability of kpc-scale observations of MaNGA, we decompose galaxies into H II, composite, and AGN-dominated regions by using the BPT diagrams. With increasing $R_{\rm mol}$, the fraction of H II regions within 1.5 effective radius decreases slightly; the density distribution in the spatially resolved BPT diagram also changes significantly, suggesting changes in metallicity and ionization states. Galaxies with high $R_{\rm mol}$ tend to have high oxygen abundance, both at one effective radius with a Kendall's $τ$ coefficient of 0.37 ($p < 10^{-3}$) and their central regions. Among all parameters investigated here, the oxygen abundance at one effective radius has the strongest relation with global $R_{\rm mol}$, but the dependence of gas conversion on gas distribution and galaxy ionization states is weak.

astro-ph.GA↗