arXiv subjects

Xin Zhang

Publications and source records attributed to Xin Zhang.

At least 505 records · Page 28Linked to original sources

arXiv2025

ReX: A Framework for Incorporating Temporal Information in Model-Agnostic Local Explanation Techniques

Existing local model-agnostic explanation techniques are ineffective for machine learning models that consider inputs of variable lengths, as they do not consider temporal information embedded in these models. To address this limitation, we propose \textsc{ReX}, a general framework for incorporating temporal information in these techniques. Our key insight is that these techniques typically learn a model surrogate by sampling model inputs and outputs, and we can incorporate temporal information in a uniform way by only changing the sampling process and the surrogate features. We instantiate our approach on three popular explanation techniques: Anchors, LIME, and Kernel SHAP. To evaluate the effectiveness of \textsc{ReX}, we apply our approach to six models in three different tasks. Our evaluation results demonstrate that our approach 1) significantly improves the fidelity of explanations, making model-agnostic techniques outperform a state-of-the-art model-specific technique on its target model, and 2) helps end users better understand the models' behaviors.

cs.LG↗

arXiv2025

Independent transversal blow-up of graphs

In an $r$-partite graph, an independent transversal of size $s$ (ITS) consists of $s$ vertices from each part forming an independent set. Motivated by a question from Bollobás, Erdős, and Szemerédi (1975), Di Braccio and Illingworth (2024) inquired about the minimum degree needed to ensure an $n \times \cdots \times n$ $r$-partite graph contains $K_r(s)$, a complete $r$-partite graph with $s$ vertices in each part. We reformulate this as finding the smallest $n$ such that any $n \times \cdots \times n$ $r$-partite graph with maximum degree $Δ$ has an ITS. For any $\varepsilon>0$, we prove the existence of a $γ>0$ ensuring that if $G$ is a multipartite graph partitioned as $(V_1, V_2, \ldots, V_r)$, where the average degree of each part $V_i$ is at most $D$, the maximum degree of any vertex to any part $V_i$ is at most $γD$, and the size of each part $V_i$ is at least $(s + \varepsilon)D$, then $G$ possesses an ITS. The constraint $(s + \varepsilon)D$ on the part size is tight. This extends results of Loh and Sudakov (2007), Glock and Sudakov (2022), and Kang and Kelly (2022). We also show that any $n \times \cdots \times n$ $r$-partite graph with minimum degree at least $\left(r-1-\frac{1}{2s^2}\right)n$ contains $K_r(s)$ and provide a relative Turán-type result. Additionally, this paper explores counting ITSs in multipartite graphs.

math.CO↗

arXiv2025

Prospects for measuring neutrino mass with 21-cm forest

Both particle physics experiments and cosmological observations have been used to explore neutrino properties. Cosmological researches of neutrinos often rely on the early-universe cosmic microwave background observations or other late-universe probes, which mostly focus on large-scale structures. We introduce a distinct probe, the 21-cm forest, that differs from other probes in both time and scale. Actually, the 21-cm forest is a unique tool for studying small-scale structures in the early universe. Below the free-streaming scale, massive neutrinos suppress the matter power spectrum, influencing small-scale fluctuations in the distribution of matter. The one-dimensional (1D) power spectrum of the 21-cm forest can track these fluctuations across different scales, similar to the matter power spectrum, providing an effective method to constrain neutrino mass. Although heating effects in the early universe can also impact the 1D power spectrum of the 21-cm forest, we assess the potential of the 21-cm forest as a tool for measuring neutrino mass, given that the temperature of the intergalactic medium can be constrained using other methods within a certain range. In the ideal scenario, the 21-cm forest observation will have the ability to constrain the total neutrino mass to around 0.1 eV. With the accumulation of observational data and advancements in observational technology, the 21-cm forest holds great promise as an emerging and potent tool for measuring neutrino mass.

astro-ph.CO↗

arXiv2025

A Novel Spike Transformer Network for Depth Estimation from Event Cameras via Cross-modality Knowledge Distillation

Depth estimation is a critical task in computer vision, with applications in autonomous navigation, robotics, and augmented reality. Event cameras, which encode temporal changes in light intensity as asynchronous binary spikes, offer unique advantages such as low latency, high dynamic range, and energy efficiency. However, their unconventional spiking output and the scarcity of labelled datasets pose significant challenges to traditional image-based depth estimation methods. To address these challenges, we propose a novel energy-efficient Spike-Driven Transformer Network (SDT) for depth estimation, leveraging the unique properties of spiking data. The proposed SDT introduces three key innovations: (1) a purely spike-driven transformer architecture that incorporates spike-based attention and residual mechanisms, enabling precise depth estimation with minimal energy consumption; (2) a fusion depth estimation head that combines multi-stage features for fine-grained depth prediction while ensuring computational efficiency; and (3) a cross-modality knowledge distillation framework that utilises a pre-trained vision foundation model (DINOv2) to enhance the training of the spiking network despite limited data availability.This work represents the first exploration of transformer-based spiking neural networks for depth estimation, providing a significant step forward in energy-efficient neuromorphic computing for real-world vision applications.

cs.CV↗

arXiv2025

Positive and Unlabeled Data: Model, Estimation, Inference, and Classification

This study introduces a new approach to addressing positive and unlabeled (PU) data through the double exponential tilting model (DETM). Traditional methods often fall short because they only apply to selected completely at random (SCAR) PU data, where the labeled positive and unlabeled positive data are assumed to be from the same distribution. In contrast, our DETM's dual structure effectively accommodates the more complex and underexplored selected at random PU data, where the labeled and unlabeled positive data can be from different distributions. We rigorously establish the theoretical foundations of DETM, including identifiability, parameter estimation, and asymptotic properties. Additionally, we move forward to statistical inference by developing a goodness-of-fit test for the SCAR condition and constructing confidence intervals for the proportion of positive instances in the target domain. We leverage an approximated Bayes classifier for classification tasks, demonstrating DETM's robust performance in prediction. Through theoretical insights and practical applications, this study highlights DETM as a comprehensive framework for addressing the challenges of PU data.

stat.ME↗

arXiv2025

DiffChip: Thermally Aware Chip Placement with Automatic Differentiation

Chiplets are modular integrated circuits that can be combined to form a larger system, offering flexibility and performance enhancements. However, their dense packing often leads to significant thermal management challenges, requiring careful floorplanning to ensure efficient heat distribution. To address thermal considerations, layout optimization algorithms concurrently minimize the total wirelength and the maximum temperature. However, these efforts employ gradient-free approaches, such as simulated annealing, which suffer from poor scaling and slow convergence. In this paper, we propose DiffChip, a chiplet placement algorithm based on automatic differentiation (AD). The proposed framework relies on a differentiable thermal solver that computes the sensitivity of the temperature map with respect to the positions of the chiplets. Regularization strategies for peak temperature, heat sources, and material properties enable end-to-end differentiability, allowing for gradient-based optimization. We apply DiffChip to optimize a layout where the total wirelength is minimized while keeping the maximum temperature below a desired threshold. By leveraging AD and physics-aware optimization, our approach accelerates the design process of microelectronic systems, exceeding traditional trial-and-error and gradient-free methods.

physics.comp-ph↗

arXiv2025

BPO: Towards Balanced Preference Optimization between Knowledge Breadth and Depth in Alignment

Reinforcement Learning with Human Feedback (RLHF) is the key to the success of large language models (LLMs) in recent years. In this work, we first introduce the concepts of knowledge breadth and knowledge depth, which measure the comprehensiveness and depth of an LLM or knowledge source respectively. We reveal that the imbalance in the number of prompts and responses can lead to a potential disparity in breadth and depth learning within alignment tuning datasets by showing that even a simple uniform method for balancing the number of instructions and responses can lead to significant improvements. Building on this, we further propose Balanced Preference Optimization (BPO), designed to dynamically augment the knowledge depth of each sample. BPO is motivated by the observation that the usefulness of knowledge varies across samples, necessitating tailored learning of knowledge depth. To achieve this, we introduce gradient-based clustering, estimating the knowledge informativeness and usefulness of each augmented sample based on the model's optimization direction. Our experimental results across various benchmarks demonstrate that BPO outperforms other baseline methods in alignment tuning while maintaining training efficiency. Furthermore, we conduct a detailed analysis of each component of BPO, providing guidelines for future research in preference data optimization.

cs.CL↗

arXiv2025

The FAST HI 21-cm absorption blind survey. II. -- Statistic Exploration for Associated and Intervening systems

We present an extragalactic HI 21-cm absorption lines catalog from a blind search at z $\leqslant$ 0.35, using drift-scan data collected in 1325.6 hours by the ongoing Commensal Radio Astronomy FasT Survey (CRAFTS) and FAST All Sky HI Survey (FASHI), which spans a sky area of 6072.0 deg$^{2}$ and covers 84533 radio sources with a flux density greater than 12 mJy. 14 previously identified HI absorbers and 20 newly discovered HI absorbers were detected, comprising 15 associated systems, 10 intervening systems, and 9 systems with undetermined classifications. Through spectral stacking, the mean peak optical path, mean velocity-integrated optical path, mean FWHM and mean HI column density are measured to be 0.47 and 0.30; 27.19 and 4.36 km s$^{-1}$; 42.61 and 9.33 km s$^{-1}$; 0.49 and 0.08 T$_{s} \times$ 10$^{20}$cm$^{-2}$K$^{-1}$, for the associated and intervening samples, respectively. Statistical analysis also reveals that associated systems tend to be hosted by red ($g-r>$0.7) galaxies at lower redshifts, whereas galaxies hosting intervening HI absorption are typically found at higher redshifts and are of a bluer ($g-r\leqslant$0.7) type. A noticeable difference is observed in the positions of foregrounds, backgrounds of intervening systems, and high-redshift and low-redshift associated systems on the WISE color-color diagram. All identified foreground sources in our sample have W1-W2 magnitudes below 0.8, suggesting no Active Galactic Nuclei (AGN). In contrast, backgrounds of intervening systems tend to have W1-W2 magnitudes above 0.8, indicating AGN presence. For associated absorption, most low-redshift ($z\leqslant$0.5) systems show W1-W2 values below 0.8, while higher-redshift associated absorption ($z>$0.5) displays a broader range of W1-W2 values.

astro-ph.GA↗

arXiv2025

Regularization by Neural Style Transfer for MRI Field-Transfer Reconstruction with Limited Data

Recent advances in MRI reconstruction have demonstrated remarkable success through deep learning-based models. However, most existing methods rely heavily on large-scale, task-specific datasets, making reconstruction in data-limited settings a critical yet underexplored challenge. While regularization by denoising (RED) leverages denoisers as priors for reconstruction, we propose Regularization by Neural Style Transfer (RNST), a novel framework that integrates a neural style transfer (NST) engine with a denoiser to enable magnetic field-transfer reconstruction. RNST generates high-field-quality images from low-field inputs without requiring paired training data, leveraging style priors to address limited-data settings. Our experiment results demonstrate RNST's ability to reconstruct high-quality images across diverse anatomical planes (axial, coronal, sagittal) and noise levels, achieving superior clarity, contrast, and structural fidelity compared to lower-field references. Crucially, RNST maintains robustness even when style and content images lack exact alignment, broadening its applicability in clinical environments where precise reference matches are unavailable. By combining the strengths of NST and denoising, RNST offers a scalable, data-efficient solution for MRI field-transfer reconstruction, demonstrating significant potential for resource-limited settings.

cs.CV↗

arXiv2025

Towards Text-Image Interleaved Retrieval

Current multimodal information retrieval studies mainly focus on single-image inputs, which limits real-world applications involving multiple images and text-image interleaved content. In this work, we introduce the text-image interleaved retrieval (TIIR) task, where the query and document are interleaved text-image sequences, and the model is required to understand the semantics from the interleaved context for effective retrieval. We construct a TIIR benchmark based on naturally interleaved wikiHow tutorials, where a specific pipeline is designed to generate interleaved queries. To explore the task, we adapt several off-the-shelf retrievers and build a dense baseline by interleaved multimodal large language model (MLLM). We then propose a novel Matryoshka Multimodal Embedder (MME), which compresses the number of visual tokens at different granularity, to address the challenge of excessive visual tokens in MLLM-based TIIR models. Experiments demonstrate that simple adaption of existing models does not consistently yield effective results. Our MME achieves significant improvements over the baseline by substantially fewer visual tokens. We provide extensive analysis and will release the dataset and code to facilitate future research.

cs.CL↗

arXiv2025

CCJA: Context-Coherent Jailbreak Attack for Aligned Large Language Models

Despite explicit alignment efforts for large language models (LLMs), they can still be exploited to trigger unintended behaviors, a phenomenon known as "jailbreaking." Current jailbreak attack methods mainly focus on discrete prompt manipulations targeting closed-source LLMs, relying on manually crafted prompt templates and persuasion rules. However, as the capabilities of open-source LLMs improve, ensuring their safety becomes increasingly crucial. In such an environment, the accessibility of model parameters and gradient information by potential attackers exacerbates the severity of jailbreak threats. To address this research gap, we propose a novel \underline{C}ontext-\underline{C}oherent \underline{J}ailbreak \underline{A}ttack (CCJA). We define jailbreak attacks as an optimization problem within the embedding space of masked language models. Through combinatorial optimization, we effectively balance the jailbreak attack success rate with semantic coherence. Extensive evaluations show that our method not only maintains semantic consistency but also surpasses state-of-the-art baselines in attack effectiveness. Additionally, by integrating semantically coherent jailbreak prompts generated by our method into widely used black-box methodologies, we observe a notable enhancement in their success rates when targeting closed-source commercial LLMs. This highlights the security threat posed by open-source LLMs to commercial counterparts. We will open-source our code if the paper is accepted.

cs.CR↗

arXiv2025

Enhancing Large Language Model Performance with Gradient-Based Parameter Selection

Large language models (LLMs) have revolutionized lots of fields of research. Although it is well-known that fine-tuning is essential for enhancing the capabilities of LLMs, existing research suggests that there is potential redundancy in the fine-tuning process and therefore proposes to update only a subset of parameters. However, these methods fail to leverage the task-specific information to identify important parameters during training. Based on the insight that gradients inherently contain information on task-specific data, we propose Gradient-Mask Tuning (GMT), a method that selectively updates parameters during training based on their gradient information. Specifically, we compute the absolute values of the gradients and apply masking to those with relatively smaller magnitudes. Our empirical results across various tasks demonstrate that GMT not only outperforms traditional fine-tuning methods but also elevates the upper limits of LLM performance. Further analysis indicates that GMT exhibits insensitivity to mask ratio and possesses computational efficiency comparable to vanilla SFT.

cs.AI↗

arXiv2025

Exploring the Key Features of Repeating Fast Radio Bursts with Machine Learning

Fast radio bursts (FRBs) are enigmatic high-energy events with unknown origins, which are observationally divided into two categories, i.e., repeaters and non-repeaters. However, there are potentially a number of non-repeaters that may be misclassified, as repeating bursts are missed due to the limited sensitivity and observation periods, thus misleading the investigation of their physical properties. In this work, we propose a repeater identification method based on the t-distributed Stochastic Neighbor Embedding (t-SNE) algorithm and apply the classification to the first Canadian Hydrogen Intensity Mapping Experiment Fast Radio Burst (CHIME/FRB) catalog. We find that the spectral morphology parameters, specifically spectral running ($r$), represent the key features for identifying repeaters from the non-repeaters. Also, the results suggest that repeaters are more biased towards narrowband emission, whereas non-repeaters are inclined toward broadband emission. We provide a list of 163 repeater candidates, 5 of which are confirmed with an updated repeater catalog from CHIME/FRB. Our findings improve our understanding of the various properties underlying repeaters and non-repeaters, as well as guidelines for future FRB detection and categorization.

astro-ph.HE↗

arXiv2025

Dynamical Models of the Milky Way in Action Space with LAMOST DR8 and GAIA EDR3

This work explores dynamical models of the Milky Way (MW) by analyzing a sample of 86,109 K giant stars selected through cross-matching the LAMOST DR8 and Gaia EDR3 surveys. Our earlier torus models in Wang et al. (2017) did not include Gaia data, making them incompatible with the new proper motion distributions of samples. Here, we refine the construction of action-based, self-consistent models to constrain the three-dimensional velocity distribution of K giants over a larger parameter space, drawing on a series of existing MW models. This approach produces several new MW models. Our best-fit model for the local kinematics near the Sun indicates a MW virial mass of 1.35 $\times 10^{12} M_\odot$, a local stellar density of 0.0696 $\rm M_\odot pc^{-3}$, and a local dark matter density of 0.0115 $\rm M_\odot pc^{-3}$. Our main conclusion supports a thicker and more extended thick disk, alongside a cooler thin disk, compared to the best-fitting model in Wang et al. (2017). Near the Sun, our model aligns well with observations, but is less satisfactory at distances far from the Galactic center, perhaps implying unidentified structures. Further high-precision observations will be critical for understanding the dynamics in these outer Galactic regions, and will require a more realistic model.

astro-ph.GA↗

arXiv2025

Non-linear integral equations for the XXX spin-1/2 quantum chain with non-diagonal boundary fields

The XXX spin-$\frac{1}{2}$ Heisenberg chain with non-diagonal boundary fields represents a cornerstone model in the study of integrable systems with open boundaries. Despite its significance, solving this model exactly has remained a formidable challenge due to the breaking of $U(1)$ symmetry. Building on the off-diagonal Bethe Ansatz (ODBA), we derive a set of nonlinear integral equations (NLIEs) that encapsulate the exact spectrum of the model. For $U(1)$ symmetric spin-$\frac{1}{2}$ chains such NLIEs involve two functions $a(x)$ and $\bar{a}(x)$ coupled by an integration kernel with short-ranged elements. The solution functions show characteristic features for arguments at some length scale which grows logarithmically with system size $N$. For the non $U(1)$ symmetric case, the equations involve a novel third function $c(x)$, which captures the inhomogeneous contributions of the $T$-$Q$ relation. The kernel elements coupling this function to the standard ones are long-ranged and lead for the ground-state to a winding phenomenon. In $\log(1+a(x))$ and $\log(1+\bar a(x))$ we observe a sudden change by $2π$i at a characteristic scale $x_1$ of the argument. Other features appear at a value $x_0$ which is of order $\log N$. These two length scales, $x_1$ and $x_0$, are independent: their ratio $x_1/x_0$ is large for small $N$ and small for large $N$. Explicit solutions to the NLIEs are obtained numerically for these limiting cases, though intermediate cases ($x_1/x_0 \sim 1$) present computational challenges. This work lays the foundation for studying finite-size corrections and conformal properties of other integrable spin chains with non-diagonal boundaries, opening new avenues for exploring boundary effects in quantum integrable systems.

cond-mat.str-el↗

arXiv2025

Magnetic Resonance Image Processing Transformer for General Accelerated Image Reconstruction

Recent advancements in deep learning have enabled the development of generalizable models that achieve state-of-the-art performance across various imaging tasks. Vision Transformer (ViT)-based architectures, in particular, have demonstrated strong feature extraction capabilities when pre-trained on large-scale datasets. In this work, we introduce the Magnetic Resonance Image Processing Transformer (MR-IPT), a ViT-based framework designed to enhance the generalizability and robustness of accelerated MRI reconstruction. Unlike conventional deep learning models that require separate training for different acceleration factors, MR-IPT is pre-trained on a large-scale dataset encompassing multiple undersampling patterns and acceleration settings, enabling a unified reconstruction framework. By leveraging a shared transformer backbone, MR-IPT effectively learns universal feature representations, allowing it to generalize across diverse reconstruction tasks. Extensive experiments demonstrate that MR-IPT outperforms both CNN-based and existing transformer-based methods, achieving superior reconstruction quality across varying acceleration factors and sampling masks. Moreover, MR-IPT exhibits strong robustness, maintaining high performance even under unseen acquisition setups, highlighting its potential as a scalable and efficient solution for accelerated MRI. Our findings suggest that transformer-based general models can significantly advance MRI reconstruction, offering improved adaptability and stability compared to traditional deep learning approaches.

eess.IV↗

arXiv2025

Pseudo Transitions in the Finite-Size Blume-Capel Model

This article investigates the pseudo transitions of the Blume-Capel model on two-dimensional finite-size lattices. By employing the Wang-Landau sampling method and microcanonical inflection point analysis, we identified the positions of phase transitions as well as higher-order phase transitions. Through Metropolis sampling and canonical ensemble analysis, we determined the geometric characteristics of the system at these transition points. When the crystal field parameter $D$ exceeds 1.965, crossing the tricritical point, no third-order dependent phase transition is observed. However, a fourth-order independent transition was identified in the high-temperature region, and through Metropolis sampling analysis, a phase transition from the ordered paramagnetic phase to the disordered paramagnetic phase was confirmed, enhancing the phase diagram. Furthermore, the positions of the third-order phase transition obtained from both microcanonical and canonical analyses are consistent and mutually corroborative. We speculate that third-order dependent transitions vanish in the presence of strong first-order phase transitions.

cond-mat.stat-mech↗

arXiv2025

Enhancing Reasoning to Adapt Large Language Models for Domain-Specific Applications

This paper presents SOLOMON, a novel Neuro-inspired Large Language Model (LLM) Reasoning Network architecture that enhances the adaptability of foundation models for domain-specific applications. Through a case study in semiconductor layout design, we demonstrate how SOLOMON enables swift adaptation of general-purpose LLMs to specialized tasks by leveraging Prompt Engineering and In-Context Learning techniques. Our experiments reveal the challenges LLMs face in spatial reasoning and applying domain knowledge to practical problems. Results show that SOLOMON instances significantly outperform their baseline LLM counterparts and achieve performance comparable to state-of-the-art reasoning model, o1-preview. We discuss future research directions for developing more adaptive AI systems that can continually learn, adapt, and evolve in response to new information and changing requirements.

cs.CL↗