arXiv subjects

Sijia Li

Publications and source records attributed to Sijia Li.

At least 55 records · Page 3Linked to original sources

arXiv2025

Efficient Inference for Covariate-adjusted Bradley-Terry Model with Covariate Shift

We propose a general framework for statistical inference on the overall strengths of players in pairwise comparisons, allowing for potential shifts in the covariate distribution. These covariates capture important contextual information that may impact the winning probability of each player. We measure the overall strengths of players under a target distribution through its Kullback-Leibler projection onto a class of covariate-adjusted Bradley-Terry model. Consequently, our estimands remain well-defined without requiring stringent model assumptions. We develop semiparametric efficient estimators and corresponding inferential procedures that allow for flexible estimation of the nuisance functions. When the conditional Bradley-Terry assumption holds, we propose additional estimators that do not require observing all pairwise comparisons. We demonstrate the performance of our proposed method in simulation studies and apply it to assess the alignment of large language models with human preferences in real-world applications.

stat.ME↗

arXiv2025

LAPTOP-Diff: Layer Pruning and Normalized Distillation for Compressing Diffusion Models

In the era of AIGC, the demand for low-budget or even on-device applications of diffusion models emerged. In terms of compressing the Stable Diffusion models (SDMs), several approaches have been proposed, and most of them leveraged the handcrafted layer removal methods to obtain smaller U-Nets, along with knowledge distillation to recover the network performance. However, such a handcrafting manner of layer removal is inefficient and lacks scalability and generalization, and the feature distillation employed in the retraining phase faces an imbalance issue that a few numerically significant feature loss terms dominate over others throughout the retraining process. To this end, we proposed the layer pruning and normalized distillation for compressing diffusion models (LAPTOP-Diff). We, 1) introduced the layer pruning method to compress SDM's U-Net automatically and proposed an effective one-shot pruning criterion whose one-shot performance is guaranteed by its good additivity property, surpassing other layer pruning and handcrafted layer removal methods, 2) proposed the normalized feature distillation for retraining, alleviated the imbalance issue. Using the proposed LAPTOP-Diff, we compressed the U-Nets of SDXL and SDM-v1.5 for the most advanced performance, achieving a minimal 4.0% decline in PickScore at a pruning ratio of 50% while the comparative methods' minimal PickScore decline is 8.2%.

cs.CV↗

arXiv2025

5G Direct Position Estimation for Precise Localization in Dense Urban Area

In recent years, the fifth-generation (5G) new radio (NR) signals have emerged as a promising supplementary resource for urban navigation. However, a major challenge in utilizing 5G signals lies in their vulnerability to non-line-of-sight (NLoS) propagation effects, which are especially prevalent in urban street canyons. This paper applies the direct position estimation (DPE) method to 5G cellular signals to mitigate the NLoS bias as well as the multipath effects, thereby enabling precise localization in urbanized environments. The feasibility of applying the DPE method to NR positioning is analyzed, followed by a discussion of the tapped delay line (TDL) channel propagation model provided by the 3rd Generation Partnership Project (3GPP). The positioning performance is then evaluated through large-scale system-level simulations. The simulation results demonstrate that 5G DPE achieves satisfactory positioning accuracy in a 10 dB noisy channel, with an overall root mean square error (RMSE) constrained within 6 m. In addition, 5G DPE outperforms the observed time difference of arrival (OTDoA) method by 95.24% in terms of positioning accuracy in an NLoS-dominated propagation environment.

eess.SP↗

arXiv2025

Efficient Estimation Under Data Fusion

We aim to make inferences about a smooth, finite-dimensional parameter by fusing data from multiple sources together. Previous works have studied the estimation of a variety of parameters in similar data fusion settings, including in the estimation of the average treatment effect and average reward under a policy, with the majority of them merging one historical data source with covariates, actions, and rewards and one data source of the same covariates. In this work, we consider the general case where one or more data sources align with each part of the distribution of the target population, for example, the conditional distribution of the reward given actions and covariates. We describe potential gains in efficiency that can arise from fusing these data sources together in a single analysis, which we characterize by a reduction in the semiparametric efficiency bound. We also provide a general means to construct estimators that achieve these bounds. In numerical experiments, we illustrate marked improvements in efficiency from using our proposed estimators rather than their natural alternatives. Finally, we illustrate the magnitude of efficiency gains that can be realized in vaccine immunogenicity studies by fusing data from two HIV vaccine trials.

stat.ME↗

arXiv2024

Color profiles of disk galaxies at $z=1$-$3$ observed with JWST: Implications for outer-disk formation histories

We investigate the deconvolved color profiles of 223 disk galaxies at redshifts of $z=1$-3 observed by the James Webb Space Telescope (JWST) as part of the Cosmic Evolution Early Release Science survey (CEERS). The filters were selected to approximate the rest-frame $B-Y$ color, which is used to identify U-shaped color profiles -- those becoming progressively bluer with increasing radius, then turning redder beyond a specific point. We find that 36% of Type II (down-bending) disks exhibit U-shaped color profiles with a minimum at or near the disk break. In contrast, no Type I (single-exponential) disks and only 9% of Type III (up-bending) disks show such a profile. The presence of U-shaped color profiles in Type II disks likely arises from the interplay between a star-formation threshold and spiral- or bar-driven secular radial migration of older stars outward. The fraction of Type II disks exhibiting a U-shaped color profile remains almost consistent across two redshift bins, $z=1$-$2$ and $z=2$-$3$, but is significantly lower than that observed in the local Universe, likely because the secular process of radial migration at high redshift may not have had sufficient time to significantly influence the disk structure. The absence of U-shaped color profiles in Type II disks could point to rapid rather than secular radial star migration potentially caused by violent clump instabilities, transporting both younger and older stars to the outer disk. Our results provide useful constraints on the formation and evolution models of disk galaxies in the early Universe.

astro-ph.GA↗

arXiv2024

Early Results from GLASS-JWST. XXV. Electron Density in the Interstellar Medium at $0.7\lesssim z\lesssim 9.3$ with NIRSpec High-resolution Spectroscopy

The electron density (${n_{\rm e}}$) of the interstellar medium (ISM) in star-forming galaxies is intimately linked to star formation and ionization condition. Using the high-resolution spectra obtained from the JWST NIRSpec micro shutter assembly (MSA) as part of the GLASS-JWST program, we have assembled the largest sample to date (34 galaxies) with individual ${n_{\rm e}}$ measurements derived from the [OII] $λλ$3726,29 and/or [SII] $λλ$6718,32 doublets at $0.7\lesssim z\lesssim 9.3$. The gravitational lensing magnification by the foreground Abell~2744 cluster allows us to probe ${n_{\rm e}}$ in galaxies with stellar masses ($M_{*}$) down to $\simeq 10^{7.5} M_\odot$ across the entire redshift range. Our analysis reveals that the [OII] flux ratios are marginally anti-correlated with specific star formation rate (sSFR) within a 1-$σ$ confidence interval, whereas the [SII] flux ratios show no significant correlation with sSFR. Despite clear correlation between sSFR and redshift within our sample, we find no apparent redshift evolution of ${n_{\rm e}}$ at $z \simeq 1-9$. Our dataset also includes 13 galaxies where ${n_{\rm e}}$ can be measured from both [OII] and [SII]. Contrary to findings at lower redshifts, we observe considerable scatter in ${n_{\rm e}}$ measurements from [OII] and [SII], indicating a complex gaseous environment with significant variations in ${n_{\rm e}}$ in high-redshift galaxies. This work highlights the unique capability of JWST NIRSpec/MSA high-resolution spectroscopy to characterize the detailed physical properties of the ISM in individual high-redshift galaxies.

astro-ph.GA↗

arXiv2024

CompoNeRF: Text-guided Multi-object Compositional NeRF with Editable 3D Scene Layout

Text-to-3D form plays a crucial role in creating editable 3D scenes for AR/VR. Recent advances have shown promise in merging neural radiance fields (NeRFs) with pre-trained diffusion models for text-to-3D object generation. However, one enduring challenge is their inadequate capability to accurately parse and regenerate consistent multi-object environments. Specifically, these models encounter difficulties in accurately representing quantity and style prompted by multi-object texts, often resulting in a collapse of the rendering fidelity that fails to match the semantic intricacies. Moreover, amalgamating these elements into a coherent 3D scene is a substantial challenge, stemming from generic distribution inherent in diffusion models. To tackle the issue of 'guidance collapse' and further enhance scene consistency, we propose a novel framework, dubbed CompoNeRF, by integrating an editable 3D scene layout with object-specific and scene-wide guidance mechanisms. It initiates by interpreting a complex text into the layout populated with multiple NeRFs, each paired with a corresponding subtext prompt for precise object depiction. Next, a tailored composition module seamlessly blends these NeRFs, promoting consistency, while the dual-level text guidance reduces ambiguity and boosts accuracy. Noticeably, our composition design permits decomposition. This enables flexible scene editing and recomposition into new scenes based on the edited layout or text prompts. Utilizing the open-source Stable Diffusion model, CompoNeRF generates multi-object scenes with high fidelity. Remarkably, our framework achieves up to a \textbf{54\%} improvement by the multi-view CLIP score metric. Our user study indicates that our method has significantly improved semantic accuracy, multi-view consistency, and individual recognizability for multi-object scene generation.

cs.CV↗

arXiv2024

Lagrangian space remapping and the angular momentum reconstruction from cosmic structures

Large scale structures provide valuable information of the primordial perturbations that encode the secrets of the origin of the Universe. It is an essential step to map between observables and their initial coordinates, called Lagrangian space, from which primordial perturbations transfer their information to structures via linear theory. By using numerical simulations and state-of-the-art reconstruction techniques, we report the accuracy of estimating the Lagrangian coordinates of galaxies and galaxy clusters, represented by dark matter halos in various ranges of mass, and study the accuracy of this remapping on the angular momentum (spin) reconstruction. Our work shows that galaxy groups and clusters, represented by halos with mass $\gtrsim 10^{13}M_\odot$, can be accurately remapped to Lagrangian space, and their spin reconstruction errors are dominated by the reconstructed initial gravitational potential. For all mass ranges, the errors of Lagrangian remapping, as well as redshift space distortions, play subdominant roles in estimating their angular momenta. This study explains the low correlation level between observed galaxy spins and reconstructed cosmic initial conditions and illustrates the potential of using angular momenta of cosmic structures to improve the reconstruction of primordial perturbations.

astro-ph.CO↗

arXiv2024

A survey on fairness of large language models in e-commerce: progress, application, and challenge

This survey explores the fairness of large language models (LLMs) in e-commerce, examining their progress, applications, and the challenges they face. LLMs have become pivotal in the e-commerce domain, offering innovative solutions and enhancing customer experiences. This work presents a comprehensive survey on the applications and challenges of LLMs in e-commerce. The paper begins by introducing the key principles underlying the use of LLMs in e-commerce, detailing the processes of pretraining, fine-tuning, and prompting that tailor these models to specific needs. It then explores the varied applications of LLMs in e-commerce, including product reviews, where they synthesize and analyze customer feedback; product recommendations, where they leverage consumer data to suggest relevant items; product information translation, enhancing global accessibility; and product question and answer sections, where they automate customer support. The paper critically addresses the fairness challenges in e-commerce, highlighting how biases in training data and algorithms can lead to unfair outcomes, such as reinforcing stereotypes or discriminating against certain groups. These issues not only undermine consumer trust, but also raise ethical and legal concerns. Finally, the work outlines future research directions, emphasizing the need for more equitable and transparent LLMs in e-commerce. It advocates for ongoing efforts to mitigate biases and improve the fairness of these systems, ensuring they serve diverse global markets effectively and ethically. Through this comprehensive analysis, the survey provides a holistic view of the current landscape of LLMs in e-commerce, offering insights into their potential and limitations, and guiding future endeavors in creating fairer and more inclusive e-commerce environments.

cs.CL↗

arXiv2024

MoEController: Instruction-based Arbitrary Image Manipulation with Mixture-of-Expert Controllers

Diffusion-model-based text-guided image generation has recently made astounding progress, producing fascinating results in open-domain image manipulation tasks. Few models, however, currently have complete zero-shot capabilities for both global and local image editing due to the complexity and diversity of image manipulation tasks. In this work, we propose a method with a mixture-of-expert (MOE) controllers to align the text-guided capacity of diffusion models with different kinds of human instructions, enabling our model to handle various open-domain image manipulation tasks with natural language instructions. First, we use large language models (ChatGPT) and conditional image synthesis models (ControlNet) to generate a large number of global image transfer dataset in addition to the instruction-based local image editing dataset. Then, using an MOE technique and task-specific adaptation training on a large-scale dataset, our conditional diffusion model can edit images globally and locally. Extensive experiments demonstrate that our approach performs surprisingly well on various image manipulation tasks when dealing with open-domain images and arbitrary human instructions. Please refer to our project page: [https://oppo-mente-lab.github.io/moe_controller/]

cs.CV↗

arXiv2024

Adversarial Training with OCR Modality Perturbation for Scene-Text Visual Question Answering

Scene-Text Visual Question Answering (ST-VQA) aims to understand scene text in images and answer questions related to the text content. Most existing methods heavily rely on the accuracy of Optical Character Recognition (OCR) systems, and aggressive fine-tuning based on limited spatial location information and erroneous OCR text information often leads to inevitable overfitting. In this paper, we propose a multimodal adversarial training architecture with spatial awareness capabilities. Specifically, we introduce an Adversarial OCR Enhancement (AOE) module, which leverages adversarial training in the embedding space of OCR modality to enhance fault-tolerant representation of OCR texts, thereby reducing noise caused by OCR errors. Simultaneously, We add a Spatial-Aware Self-Attention (SASA) mechanism to help the model better capture the spatial relationships among OCR tokens. Various experiments demonstrate that our method achieves significant performance improvements on both the ST-VQA and TextVQA datasets and provides a novel paradigm for multimodal adversarial training.

cs.CV↗

arXiv2024

Risk Taxonomy, Mitigation, and Assessment Benchmarks of Large Language Model Systems

Large language models (LLMs) have strong capabilities in solving diverse natural language processing tasks. However, the safety and security issues of LLM systems have become the major obstacle to their widespread application. Many studies have extensively investigated risks in LLM systems and developed the corresponding mitigation strategies. Leading-edge enterprises such as OpenAI, Google, Meta, and Anthropic have also made lots of efforts on responsible LLMs. Therefore, there is a growing need to organize the existing studies and establish comprehensive taxonomies for the community. In this paper, we delve into four essential modules of an LLM system, including an input module for receiving prompts, a language model trained on extensive corpora, a toolchain module for development and deployment, and an output module for exporting LLM-generated content. Based on this, we propose a comprehensive taxonomy, which systematically analyzes potential risks associated with each module of an LLM system and discusses the corresponding mitigation strategies. Furthermore, we review prevalent benchmarks, aiming to facilitate the risk assessment of LLM systems. We hope that this paper can help LLM participants embrace a systematic perspective to build their responsible LLM systems.

cs.CL↗

arXiv2023

More than Correlation: Do Large Language Models Learn Causal Representations of Space?

Recent work found high mutual information between the learned representations of large language models (LLMs) and the geospatial property of its input, hinting an emergent internal model of space. However, whether this internal space model has any causal effects on the LLMs' behaviors was not answered by that work, led to criticism of these findings as mere statistical correlation. Our study focused on uncovering the causality of the spatial representations in LLMs. In particular, we discovered the potential spatial representations in DeBERTa, GPT-Neo using representational similarity analysis and linear and non-linear probing. Our casual intervention experiments showed that the spatial representations influenced the model's performance on next word prediction and a downstream task that relies on geospatial information. Our experiments suggested that the LLMs learn and use an internal model of space in solving geospatial related tasks.

cs.CL↗

arXiv2023

Big Learning Expectation Maximization

Mixture models serve as one fundamental tool with versatile applications. However, their training techniques, like the popular Expectation Maximization (EM) algorithm, are notoriously sensitive to parameter initialization and often suffer from bad local optima that could be arbitrarily worse than the optimal. To address the long-lasting bad-local-optima challenge, we draw inspiration from the recent ground-breaking foundation models and propose to leverage their underlying big learning principle to upgrade the EM. Specifically, we present the Big Learning EM (BigLearn-EM), an EM upgrade that simultaneously performs joint, marginal, and orthogonally transformed marginal matchings between data and model distributions. Through simulated experiments, we empirically show that the BigLearn-EM is capable of delivering the optimal with high probability; comparisons on benchmark clustering datasets further demonstrate its effectiveness and advantages over existing techniques. The code is available at https://github.com/YulaiCong/Big-Learning-Expectation-Maximization.

cs.LG↗

arXiv2023

Future Full-Ocean Deep SSPs Prediction based on Hierarchical Long Short-Term Memory Neural Networks

The spatial-temporal distribution of underwater sound velocity affects the propagation mode of underwater acoustic signals. Therefore, rapid estimation and prediction of underwater sound velocity distribution is crucial for providing underwater positioning, navigation and timing (PNT) services. Currently, sound speed profile (SSP) inversion methods have a faster time response rate compared to direct measurement methods, however, most SSP inversion methods focus on constructing spatial dimensional sound velocity fields and are highly dependent on sonar observation data, thus high requirements have been placed on observation data sources. To explore the distribution pattern of sound velocity in the time dimension and achieve future SSP prediction without sonar observation data, we propose a hierarchical long short-term memory (H-LSTM) neural network for SSP prediction. By our SSP prediction method, the sound speed distribution could be estimated without any on-site data measurement process, so that the time efficiency could be greatly improved. Through comparing with other state-of-the-art methods, H-LSTM has better accuracy performance on prediction of monthly average sound velocity distribution, which is less than 1 m/s in different depth layers.

cs.SD↗

arXiv2023

Underwater Sound Speed Profile Construction: A Review

Real--time and accurate construction of regional sound speed profiles (SSP) is important for building underwater positioning, navigation, and timing (PNT) systems as it greatly affect the signal propagation modes such as trajectory. In this paper, we summarizes and analyzes the current research status in the field of underwater SSP construction, and the mainstream methods include direct SSP measurement and SSP inversion. In the direct measurement method, we compare the performance of popular international commercial temperature, conductivity, and depth profilers (CTD). While for the inversion methods, the framework and basic principles of matched field processing (MFP), compressive sensing (CS), and deep learning (DL) for constructing SSP are introduced, and their advantages and disadvantages are compared. The traditional direct measurement method has good accuracy performance, but it usually takes a long time. The proposal of SSP inversion method greatly improves the convenience and real--time performance, but the accuracy is not as good as the direct measurement method. Currently, the SSP inversion relies on sonar observation data, making it difficult to apply to areas that couldn't be covered by underwater observation systems, and these methods are unable to predict the distribution of sound velocity at future times. How to comprehensively utilize multi-source data and provide elastic sound velocity distribution estimation services with different accuracy and real-time requirements for underwater users without sonar observation data is the mainstream trend in future research on SSP construction.

eess.SP↗

arXiv2023

Baryonic Effects on Lagrangian Clustering and Angular Momentum Reconstruction

Recent studies illustrate the correlation between the angular momenta of cosmic structures and their Lagrangian properties. However, only baryons are observable and it is unclear whether they reliably trace the cosmic angular momenta. We study the Lagrangian mass distribution, spin correlation, and predictability of dark matter, gas, and stellar components of galaxy-halo systems using IllustrisTNG, and show that the primordial segregations between components are typically small. Their protoshapes are also similar in terms of the statistics of moment of inertia tensors. Under the common gravitational potential they are expected to exert the same tidal torque and the strong spin correlations are not destroyed by the nonlinear evolution and complicated baryonic effects, as confirmed by the high-resolution hydrodynamic simulations. We further show that their late-time angular momenta traced by total gas, stars, or the central galaxies, can be reliably reconstructed by the initial perturbations. These results suggest that baryonic angular momenta can potentially be used in reconstructing the parameters and models related to the initial perturbations.

astro-ph.CO↗

arXiv2022

ScrollTest: Evaluating Scrolling Speed and Accuracy

Scrolling is an essential interaction technique enabling users to display previously off-screen content. Existing evaluation models for scrolling are often entangled with the selection of content, e.g., when scrolling on the phone for reading. Furthermore, some evaluation models overlook whether the user knows the target position. We have developed ScrollTest, a general-purpose evaluation tool for scrolling speed and accuracy that avoids the need for selection. We tested it across four dimensions: 11 different scrolling techniques/devices, 5 frame heights, 13 scrolling distances, and 2 scrolling conditions (i.e., with or without knowing the target position). The results show that flicking and two-finger scrolling are the fastest; flicking is also relatively precise for scrolling to targets already onscreen, but pressing arrow buttons on the scrollbar is the most accurate for scrolling to nearby targets. Mathematical models of scrolling are highly linear when the target position is unknown but like Fitts' law when known.

cs.HC↗