arXiv subjects

Xin Huang

Publications and source records attributed to Xin Huang.

At least 109 records · Page 6Linked to original sources

arXiv2025

Investigating and Scaling up Code-Switching for Multilingual Language Model Pre-Training

Large language models (LLMs) exhibit remarkable multilingual capabilities despite the extreme language imbalance in the pre-training data. In this paper, we closely examine the reasons behind this phenomenon, focusing on the pre-training corpus. We find that the existence of code-switching, alternating between different languages within a context, is key to multilingual capabilities. We conduct an analysis to investigate code-switching in the pre-training corpus, examining its presence and categorizing it into four types within two quadrants. We then assess its impact on multilingual performance. These types of code-switching data are unbalanced in proportions and demonstrate different effects on facilitating language transfer. To better explore the power of code-switching for language alignment during pre-training, we investigate the strategy of synthetic code-switching. We continuously scale up the synthetic code-switching data and observe remarkable improvements in both benchmarks and representation space. Extensive experiments indicate that incorporating synthetic code-switching data enables better language alignment and generalizes well to high, medium, and low-resource languages with pre-training corpora of varying qualities.

cs.CL↗

arXiv2025

Control of atomic reconstruction and quasi-1D excitons in strain-engineered moiré heterostructures

In two-dimensional nearly commensurate heterostructures, strain plays a critical role in shaping electronic behavior. While previous studies have focused on random strain introduced during fabrication, achieving controlled structural design has remained challenging. Here, we demonstrate the deterministic creation of one-dimensional arrays from initially zero-dimensional triangular moiré patterns in MoSe$_2$-WSe$_2$ heterobilayers. This transformation, driven by the interplay of uniaxial strain and atomic reconstruction, results in one-dimensional confinement of interlayer excitons within domain walls, exhibiting near-unity linearly polarized emission related to the confinement-induced symmetry breaking. The width of the domain walls--and consequently the degree of exciton confinement--can be precisely tuned by the interlayer twist angle. By applying out-of-plane electric field, the confined excitons exhibit energy shifts exceeding 100~meV and changes in the fine-structure splitting by up to a factor of two. Our work demonstrates the potential of strain engineering for constructing designer moiré systems with programmable quantum properties, paving the way for future optoelectronic applications.

cond-mat.mes-hall↗

arXiv2025

On critical maps of the horizontal energy functional between Riemannian foliations

In this paper, we consider critical points of the horizontal energy $E_{\HH}(f)$ for a smooth map $f$ between two Riemannian foliations. These critical points are referred to as horizontally harmonic maps. In particular, if the maps are foliated, they become transversally harmonic maps. By utilizing the stress-energy tensor, we establish some monotonicity formulas for horizontally harmonic maps from Euclidean spaces, the quotients $K_{m}$ of Heisenberg groups and also for transversally harmonic maps from Riemannian foliations with appropriate curvature pinching conditions. Finally, we give Jin-type theorems for either horizontally harmonic maps or transversally harmonic maps under some asymptotic conditions at infinity.

math.DG↗

arXiv2025

R-PRM: Reasoning-Driven Process Reward Modeling

Large language models (LLMs) inevitably make mistakes when performing step-by-step mathematical reasoning. Process Reward Models (PRMs) have emerged as a promising solution by evaluating each reasoning step. However, existing PRMs typically output evaluation scores directly, limiting both learning efficiency and evaluation accuracy, which is further exacerbated by the scarcity of annotated data. To address these issues, we propose Reasoning-Driven Process Reward Modeling (R-PRM). First, we leverage stronger LLMs to generate seed data from limited annotations, effectively bootstrapping our model's reasoning capabilities and enabling comprehensive step-by-step evaluation. Second, we further enhance performance through preference optimization, without requiring additional annotated data. Third, we introduce inference-time scaling to fully harness the model's reasoning potential. Extensive experiments demonstrate R-PRM's effectiveness: on ProcessBench and PRMBench, it surpasses strong baselines by 11.9 and 8.5 points in F1 scores, respectively. When applied to guide mathematical reasoning, R-PRM achieves consistent accuracy improvements of over 8.5 points across six challenging datasets. Further analysis reveals that R-PRM exhibits more comprehensive evaluation and stronger generalization capabilities, thereby highlighting its significant potential.

cs.CL↗

arXiv2025

M-DocSum: Do LVLMs Genuinely Comprehend Interleaved Image-Text in Document Summarization?

We investigate a critical yet under-explored question in Large Vision-Language Models (LVLMs): Do LVLMs genuinely comprehend interleaved image-text in the document? Existing document understanding benchmarks often assess LVLMs using question-answer formats, which are information-sparse and difficult to guarantee the coverage of long-range dependencies. To address this issue, we introduce a novel and challenging Multimodal Document Summarization Benchmark (M-DocSum-Bench), which comprises 500 high-quality arXiv papers, along with interleaved multimodal summaries aligned with human preferences. M-DocSum-Bench is a reference-based generation task and necessitates the generation of interleaved image-text summaries using provided reference images, thereby simultaneously evaluating capabilities in understanding, reasoning, localization, and summarization within complex multimodal document scenarios. To facilitate this benchmark, we develop an automated framework to construct summaries and propose a fine-grained evaluation method called M-DocEval. Moreover, we further develop a robust summarization baseline, i.e., M-DocSum-7B, by progressive two-stage training with diverse instruction and preference data. The extensive results on our M-DocSum-Bench reveal that the leading LVLMs struggle to maintain coherence and accurately integrate information within long and interleaved contexts, often exhibiting confusion between similar images and a lack of robustness. Notably, M-DocSum-7B achieves state-of-the-art performance compared to larger and closed-source models (including GPT-4o, Gemini Pro, Claude-3.5-Sonnet and Qwen2.5-VL-72B, etc.), demonstrating the potential of LVLMs for improved interleaved image-text understanding. The code, data, and models are available at https://github.com/stepfun-ai/M-DocSum-Bench.

cs.CV↗

arXiv2025

EVA-MED: An Enhanced Valence-Arousal Multimodal Emotion Dataset for Emotion Recognition

We introduce a novel multimodal emotion recognition dataset that enhances the precision of Valence-Arousal Model while accounting for individual differences. This dataset includes electroencephalography (EEG), electrocardiography (ECG), and pulse interval (PI) from 64 participants. Data collection employed two emotion induction paradigms: video stimuli that targeted different valence levels (positive, neutral, and negative) and the Mannheim Multicomponent Stress Test (MMST), which induced high arousal through cognitive, emotional, and social stressors. To enrich the dataset, participants' personality traits, anxiety, depression, and emotional states were assessed using validated questionnaires. By capturing a broad spectrum of affective responses while accounting for individual differences, this dataset provides a robust resource for precise emotion modeling. The integration of multimodal physiological data with psychological assessments lays a strong foundation for personalized emotion recognition. We anticipate this resource will support the development of more accurate, adaptive, and individualized emotion recognition systems across diverse applications.

cs.HC↗

arXiv2025

Whoever Said Money Won't Solve All Your Problems? Weighted Envy-free Allocation with Subsidy

We explore solutions for fairly allocating indivisible items among agents assigned weights representing their entitlements. Our fairness goal is weighted-envy-freeness (WEF), where each agent deems their allocated portion relative to their entitlement at least as favorable as any others relative to their own. Often, achieving WEF necessitates monetary transfers, which can be modeled as third-party subsidies. The goal is to attain WEF with bounded subsidies. Previous work relied on characterizations of unweighted envy-freeness (EF), that fail in the weighted setting. This makes our new setting challenging. We present polynomial-time algorithms that compute WEF allocations with a guaranteed upper bound on total subsidy for monotone valuations and various subclasses thereof. We also present an efficient algorithm to compute a fair allocation of items and money, when the budget is not enough to make the allocation WEF. This algorithm is new even for the unweighted setting.

cs.GT↗

arXiv2025

Large Language Models Are Cross-Lingual Knowledge-Free Reasoners

Large Language Models have demonstrated impressive reasoning capabilities across multiple languages. However, the relationship between capabilities in different languages is less explored. In this work, we decompose the process of reasoning tasks into two separated components: knowledge retrieval and knowledge-free reasoning, and analyze the relationship between cross-lingual transferability and these two components. With adapted commonsense reasoning datasets and constructed knowledge-free reasoning datasets, we show that the knowledge-free reasoning capability can be nearly perfectly transferred across various source-target language directions despite the secondary impact of resource in some specific target languages, while cross-lingual knowledge retrieval significantly hinders the transfer. Moreover, by analyzing the hidden states and feed-forward network neuron activation during the reasoning, we show that higher similarity of hidden representations and larger overlap of activated neurons could explain the better cross-lingual transferability of knowledge-free reasoning than knowledge retrieval. Thus, we hypothesize that knowledge-free reasoning shares similar neurons in different languages for reasoning, while knowledge is stored separately in different languages. Our code and data is available at: https://github.com/NJUNLP/Knowledge-Free-Reasoning.

cs.CL↗

arXiv2025

On Morita equivalences with endopermutation source and isotypies

We introduce a new type of equivalence between blocks of finite group algebras called an almost isotypy. An almost isotypy restricts to a weak isotypy in Broué's original definition, and it is slightly weaker than Linckelmann's version. We show that a bimodule of two block algebras of finite groups - which has an endopermutation module as a source and which induces a Morita equivalence - gives rise, via slash functors, to an almost isotypy if the character values of a (hence any) source are rational integers. Consequently, if two blocks are Morita equivalent via a bimodule with endopermutation source, then they are almost isotypic. We also explain why the notion of almost isotypies is reasonable.

math.RT↗

arXiv2025

Enhancing Web Service Anomaly Detection via Fine-grained Multi-modal Association and Frequency Domain Analysis

Anomaly detection is crucial for ensuring the stability and reliability of web service systems. Logs and metrics contain multiple information that can reflect the system's operational state and potential anomalies. Thus, existing anomaly detection methods use logs and metrics to detect web service systems' anomalies through data fusion approaches. They associate logs and metrics using coarse-grained time window alignment and capture the normal patterns of system operation through reconstruction. However, these methods have two issues that limit their performance in anomaly detection. First, due to asynchrony between logs and metrics, coarse-grained time window alignment cannot achieve a precise association between the two modalities. Second, reconstruction-based methods suffer from severe overgeneralization problems, resulting in anomalies being accurately reconstructed. In this paper, we propose a novel anomaly detection method named FFAD to address these two issues. On the one hand, FFAD employs graph-based alignment to mine and extract associations between the modalities from the constructed log-metric relation graph, achieving precise associations between logs and metrics. On the other hand, we improve the model's fit to normal data distributions through Fourier Frequency Focus, thereby enhancing the effectiveness of anomaly detection. We validated the effectiveness of our model on two real-world industrial datasets and one open-source dataset. The results show that our method achieves an average anomaly detection F1-score of 93.6%, representing an 8.8% improvement over previous state-of-the-art methods.

cs.SE↗

arXiv2025

MERaLiON-TextLLM: Cross-Lingual Understanding of Large Language Models in Chinese, Indonesian, Malay, and Singlish

Multilingual large language models (MLLMs) have shown impressive capabilities across a variety of languages. However, efficacy can differ greatly between different language families, especially for those with limited linguistic resources. This report presents MERaLiON-TextLLM, a series of open-source language models specifically tailored to improve understanding and generation in Chinese, Indonesian, Malay, and Singlish. The initial released model is built on Llama-3-8B-Base and refined through a meticulously crafted process of continued pre-training and weight merging. Our approach achieves performance improvements across benchmarks in these languages, exceeding the capabilities of the official Llama-3 models. We provide the model checkpoints as a resource to support further research and development in cross-lingual language understanding.

cs.CL↗

arXiv2025

On stability of exponentially subelliptic harmonic maps

In this paper, we study the stability problem of exponentially subelliptic harmonic maps from sub-Riemannian manifolds to Riemannian manifolds. We derive the rst and second variation formulas for exponentially subelliptic harmonic maps, and apply these formulas to prove that if the target manifold has nonpositive curvature, the exponentially subelliptic harmonic map is stable. Further, we obtain the instability of exponentially subelliptic harmonic maps when the target manifold is a sphere.

math.DG↗

arXiv2025

Optimization and Application of Cloud-based Deep Learning Architecture for Multi-Source Data Prediction

This study develops a cloud-based deep learning system for early prediction of diabetes, leveraging the distributed computing capabilities of the AWS cloud platform and deep learning technologies to achieve efficient and accurate risk assessment. The system utilizes EC2 p3.8xlarge GPU instances to accelerate model training, reducing training time by 93.2% while maintaining a prediction accuracy of 94.2%. With an automated data processing and model training pipeline built using Apache Airflow, the system can complete end-to-end updates within 18.7 hours. In clinical applications, the system demonstrates a prediction accuracy of 89.8%, sensitivity of 92.3%, and specificity of 95.1%. Early interventions based on predictions lead to a 37.5% reduction in diabetes incidence among the target population. The system's high performance and scalability provide strong support for large-scale diabetes prevention and management, showcasing significant public health value.

cs.DC↗

arXiv2025

LAMOST J101356.33+272410.7: A Detached White Dwarf-Main-sequence Binary with a Massive White Dwarf Within the Period Gap

We report the analysis of the detached eclipsing spectroscopic binary system LAMOST J101356.33+272410.7, which features a massive white dwarf. Using LAMOST and SDSS spectra, we determined the stellar parameters and radial velocities of both components. SED fitting of photometric data from GALEX, 2MASS, and Pan-STARRS1 yielded the effective temperatures and photometric radii. Eclipsing analysis of high-speed photometric data from the Liverpool Telescope provided orbital inclination, masses, radii, and related physical parameters. The white dwarf in this system has a mass of $1.05 \pm 0.09 \, M_\odot$ and a radius of $0.0090 \pm 0.0008 \, R_\odot$, while the main-sequence star has a mass of $0.299 \pm 0.045 \, M_\odot$ and a radius of $0.286 \pm 0.018 \, R_\odot$. Emission lines observed in the spectra indicate the likely presence of stellar magnetic activity in this system. The relatively cool temperature of the white dwarf suggests that the system could be a post-common-envelope binary (PCEB) that has not undergone mass transfer, while the presence of a massive white dwarf indicates that the system might also represent a detached cataclysmic variable (dCV) crossing the period gap. We suggest that the system is more likely to be a PCEB, and it is predicted to evolve into a cataclysmic variable and begin mass transfer in approximately 0.27 Gyr.

astro-ph.SR↗

arXiv2025

On solvmanifolds with complex commutator and constant holomorphic sectional curvature

An old open question in non-Kähler geometry predicts that any compact Hermitian manifold with constant holomorphic sectional curvature must be Kähler or Chern flat. The conjecture is known to be true in dimension $2$ due to the work by Balas-Gauduchon and Apostolov-Davidov-Muskarov in the 1980s and 1990s, but is still open in dimensions $3$ or higher, except in several special cases. The difficulty in this quest for `Hermitian space forms' is largely due to the algebraic complicity or lack of symmetry for the curvature tensor of a general Hermitian metric. In this article, we confirm the conjecture for all solvmanifolds with complex commutator, extending earlier result on nilmanifolds by Li and the second named author.

math.DG↗

arXiv2024

Bright-NeRF:Brightening Neural Radiance Field with Color Restoration from Low-light Raw Images

Neural Radiance Fields (NeRFs) have demonstrated prominent performance in novel view synthesis. However, their input heavily relies on image acquisition under normal light conditions, making it challenging to learn accurate scene representation in low-light environments where images typically exhibit significant noise and severe color distortion. To address these challenges, we propose a novel approach, Bright-NeRF, which learns enhanced and high-quality radiance fields from multi-view low-light raw images in an unsupervised manner. Our method simultaneously achieves color restoration, denoising, and enhanced novel view synthesis. Specifically, we leverage a physically-inspired model of the sensor's response to illumination and introduce a chromatic adaptation loss to constrain the learning of response, enabling consistent color perception of objects regardless of lighting conditions. We further utilize the raw data's properties to expose the scene's intensity automatically. Additionally, we have collected a multi-view low-light raw image dataset to advance research in this field. Experimental results demonstrate that our proposed method significantly outperforms existing 2D and 3D approaches. Our code and dataset will be made publicly available.

cs.CV↗

arXiv2024

Neural LightRig: Unlocking Accurate Object Normal and Material Estimation with Multi-Light Diffusion

Recovering the geometry and materials of objects from a single image is challenging due to its under-constrained nature. In this paper, we present Neural LightRig, a novel framework that boosts intrinsic estimation by leveraging auxiliary multi-lighting conditions from 2D diffusion priors. Specifically, 1) we first leverage illumination priors from large-scale diffusion models to build our multi-light diffusion model on a synthetic relighting dataset with dedicated designs. This diffusion model generates multiple consistent images, each illuminated by point light sources in different directions. 2) By using these varied lighting images to reduce estimation uncertainty, we train a large G-buffer model with a U-Net backbone to accurately predict surface normals and materials. Extensive experiments validate that our approach significantly outperforms state-of-the-art methods, enabling accurate surface normal and PBR material estimation with vivid relighting effects. Code and dataset are available on our project page at https://projects.zxhezexin.com/neural-lightrig.

cs.CV↗

arXiv2024

Style3D: Attention-guided Multi-view Style Transfer for 3D Object Generation

We present Style3D, a novel approach for generating stylized 3D objects from a content image and a style image. Unlike most previous methods that require case- or style-specific training, Style3D supports instant 3D object stylization. Our key insight is that 3D object stylization can be decomposed into two interconnected processes: multi-view dual-feature alignment and sparse-view spatial reconstruction. We introduce MultiFusion Attention, an attention-guided technique to achieve multi-view stylization from the content-style pair. Specifically, the query features from the content image preserve geometric consistency across multiple views, while the key and value features from the style image are used to guide the stylistic transfer. This dual-feature alignment ensures that spatial coherence and stylistic fidelity are maintained across multi-view images. Finally, a large 3D reconstruction model is introduced to generate coherent stylized 3D objects. By establishing an interplay between structural and stylistic features across multiple views, our approach enables a holistic 3D stylization process. Extensive experiments demonstrate that Style3D offers a more flexible and scalable solution for generating style-consistent 3D assets, surpassing existing methods in both computational efficiency and visual quality.

cs.CV↗