arXiv subjects

Hao Yu

Publications and source records attributed to Hao Yu.

At least 109 records · Page 6Linked to original sources

arXiv2025

UI-TARS-2 Technical Report: Advancing GUI Agent with Multi-Turn Reinforcement Learning

The development of autonomous agents for graphical user interfaces (GUIs) presents major challenges in artificial intelligence. While recent advances in native agent models have shown promise by unifying perception, reasoning, action, and memory through end-to-end learning, open problems remain in data scalability, multi-turn reinforcement learning (RL), the limitations of GUI-only operation, and environment stability. In this technical report, we present UI-TARS-2, a native GUI-centered agent model that addresses these challenges through a systematic training methodology: a data flywheel for scalable data generation, a stabilized multi-turn RL framework, a hybrid GUI environment that integrates file systems and terminals, and a unified sandbox platform for large-scale rollouts. Empirical evaluation demonstrates that UI-TARS-2 achieves significant improvements over its predecessor UI-TARS-1.5. On GUI benchmarks, it reaches 88.2 on Online-Mind2Web, 47.5 on OSWorld, 50.6 on WindowsAgentArena, and 73.3 on AndroidWorld, outperforming strong baselines such as Claude and OpenAI agents. In game environments, it attains a mean normalized score of 59.8 across a 15-game suite-roughly 60% of human-level performance-and remains competitive with frontier proprietary models (e.g., OpenAI o3) on LMGame-Bench. Additionally, the model can generalize to long-horizon information-seeking tasks and software engineering benchmarks, highlighting its robustness across diverse agent tasks. Detailed analyses of training dynamics further provide insights into achieving stability and efficiency in large-scale agent RL. These results underscore UI-TARS-2's potential to advance the state of GUI agents and exhibit strong generalization to real-world interactive scenarios.

cs.AI↗

arXiv2025

Barron regularity of many particle Schrödinger eigenfunctions

This work investigates the regularity of Schrödinger eigenfunctions and the solvability of Schrödinger equations in spectral Barron space $\mathcal{B}^{s}(\mathbb{R}^{nN})$, where neural networks exhibit dimension-free approximation capabilities. Under assumptions that the potential $V$ consists of one-particle and pairwise interaction parts $V_{i},V_{ij}$ in Fourier-Lebesgue space $\mathcal{F}L_{s}^{1}(\mathbb{R}^{n})+\mathcal{F}L_{s}^{α^{\prime}}(\mathbb{R}^{n})$ and an additional part $V_{\operatorname{a d}} \in \mathcal{F}L_{s}^{1}(\mathbb{R}^{nN})$, we prove that all eigenfunctions $ψ\in \bigcap_{γ 0$. The assumption accommodates many prevalent singular potentials, such as inverse power potentials. Moreover, under the same assumption or a stronger assumption $V\in\mathcal{B}^{s}(\mathbb{R}^{nN})$, we establish the solvability of Schrödinger equations and derive compactness results for $V\in\mathcal{B}^{s}(\mathbb{R}^{nN})$ with $s>-1$.

math.AP↗

arXiv2025

Polarization Images of Solitonic Boson Stars

This study investigates the polarization characteristics of solitonic boson stars surrounded by a thin accretion disk. By comparing their polarization images with corresponding optical images, we find a positive correlation between the polarization intensity distribution in the polarization images and the brightness in the optical images. Consequently, the strongest polarization occurs at the location corresponding to the direct image. The influence of the coupling strength of the sixtic potential on the polarization intensity distribution is not monotonic, under strong coupling, the polarization will concentrated on the left side of the image as the coupling strength increases, whereas under weak coupling, it is more evenly distributed across the entire direct image as the coupling strength increases. Moreover, we find that as the initial scalar field increases, both the lensing image and photon ring become more prominent. However, the polarization intensity at these regions remains weak. Due to the absence of the event horizon in solitonic boson stars, the polarization vector can penetrate the stellar interior, unlike in black holes, where no polarization signals exist within the event horizon. Our numerical simulations clearly reveal this phenomenon, suggesting that polarization features may serve as an effective tool for distinguishing solitonic boson stars from black holes.

gr-qc↗

arXiv2025

Gen-AFFECT: Generation of Avatar Fine-grained Facial Expressions with Consistent identiTy

Different forms of customized 2D avatars are widely used in gaming applications, virtual communication, education, and content creation. However, existing approaches often fail to capture fine-grained facial expressions and struggle to preserve identity across different expressions. We propose GEN-AFFECT, a novel framework for personalized avatar generation that generates expressive and identity-consistent avatars with a diverse set of facial expressions. Our framework proposes conditioning a multimodal diffusion transformer on an extracted identity-expression representation. This enables identity preservation and representation of a wide range of facial expressions. GEN-AFFECT additionally employs consistent attention at inference for information sharing across the set of generated expressions, enabling the generation process to maintain identity consistency over the array of generated fine-grained expressions. GEN-AFFECT demonstrates superior performance compared to previous state-of-the-art methods on the basis of the accuracy of the generated expressions, the preservation of the identity and the consistency of the target identity across an array of fine-grained facial expressions.

cs.CV↗

arXiv2025

Large-Small Model Collaborative Framework for Federated Continual Learning

Continual learning (CL) for Foundation Models (FMs) is an essential yet underexplored challenge, especially in Federated Continual Learning (FCL), where each client learns from a private, evolving task stream under strict data and communication constraints. Despite their powerful generalization abilities, FMs often exhibit suboptimal performance on local downstream tasks, as they are unable to utilize private local data. Furthermore, enabling FMs to learn new tasks without forgetting prior knowledge is inherently a challenging problem, primarily due to their immense parameter count and high model complexity. In contrast, small models can be trained locally under resource-constrained conditions and benefit from more mature CL techniques. To bridge the gap between small models and FMs, we propose the first collaborative framework in FCL, where lightweight local models act as a dynamic bridge, continually adapting to new tasks while enhancing the utility of the large model. Two novel components are also included: Small Model Continual Fine-tuning is for preventing small models from temporal forgetting; One-by-One Distillation performs personalized fusion of heterogeneous local knowledge on the server. Experimental results demonstrate its superior performance, even when clients utilize heterogeneous small models.

cs.LG↗

arXiv2025

Optical Characteristics of the Kerr-Bertotti-Robinson Black Hole

The Kerr-Bertotti-Robinson (Kerr-BR) black hole, a theoretical model of a rotating black hole immersed in a uniform magnetic field, has been proposed recently by Podolsky and Ovcharenko. This study investigates the optical characteristics of the Kerr-BR black hole based on the exact solution. We analyze the optical image under two illumination models: a celestial light source and a geometrically thin accretion disk. We reveal distinct roles for the fundamental parameters in the model. Specifically, it is found that under both illumination models, the effect of the rotation parameter on the optical image of the Kerr-BR black hole is significantly different from that of the magnetic field. As the magnetic field increases, the radii of both the shadow and the Einstein ring enlarge. We also attempt to use the data from M87* and Sgr A* to constrain the magnetic field. These results enhance our understanding of the optical characteristics of the Kerr-BR black hole and establish a theoretical foundation for interpreting future observations on the optical image of the black hole immersed in a uniform magnetic field. Finally, we point out that with advances in the resolution of black hole images, it is possible to detect potential BR-like magnetic fields around black holes.

gr-qc↗

arXiv2025

LLM-Barber: Block-Aware Rebuilder for Sparsity Mask in One-Shot for Large Language Models

Large language models (LLMs) have seen substantial growth, necessitating efficient model pruning techniques. Existing post-training pruning methods primarily measure weight importance in converged dense models, often overlooking changes in weight significance during the pruning process, leading to performance degradation. To address this issue, we present LLM-Barber (Block-Aware Rebuilder for Sparsity Mask in One-Shot), a novel one-shot pruning framework that rebuilds the sparsity mask of pruned models without any retraining or weight reconstruction. LLM-Barber incorporates block-aware error optimization across Self-Attention and MLP blocks, facilitating global performance optimization. We are the first to employ the product of weights and gradients as a pruning metric in the context of LLM post-training pruning. This enables accurate identification of weight importance in massive models and significantly reduces computational complexity compared to methods using secondorder information. Our experiments show that LLM-Barber efficiently prunes models from LLaMA and OPT families (7B to 13B) on a single A100 GPU in just 30 minutes, achieving state-of-the-art results in both perplexity and zero-shot performance across various language benchmarks. Code is available at https://github.com/YupengSu/LLM-Barber.

cs.LG↗

arXiv2025

Subgroup Perfect Codes of $\mathcal{A}_t$-Groups and Their Applications

A subset $C$ of the vertex set of a graph $Γ$ is called a perfect code in $Γ$ if every vertex of $Γ$ is at distance no more than 1 to exactly one vertex of $C$. A subgroup $H$ of a group $G$ is called a subgroup perfect code of $G$ if $H$ is a perfect code in some Cayley graph of $G$. Recently, Zhang reveals that the study of subgroup perfect codes of finite groups naturally reduces to the case of $p$-groups, especially $2$-groups. Based on the combined works of Berkovich, Janko and Zhang, every $p$-group is an $\mathcal{A}_t$-group. In this work, we establish a complete classification of subgroup perfect codes of $\mathcal{A}_t$-groups for $t \in\{0, 1\}$. Moreover, subgroup perfect codes of finite groups with abelian Sylow $2$-subgroups are also characterized.

math.CO↗

arXiv2025

SimdBench: Benchmarking Large Language Models for SIMD-Intrinsic Code Generation

SIMD (Single Instruction Multiple Data) instructions and their compiler intrinsics are widely supported by modern processors to accelerate performance-critical tasks. SIMD intrinsic programming, a trade-off between coding productivity and high performance, is widely used in the development of mainstream performance-critical libraries and daily computing tasks. Large Language Models (LLMs), which have demonstrated strong and comprehensive capabilities in code generation, show promise in assisting programmers with the challenges of SIMD intrinsic programming. However, existing code-generation benchmarks focus on only scalar code, and it is unclear how LLMs perform in generating vectorized code using SIMD intrinsics. To fill this gap, we propose SimdBench, the first code benchmark specifically designed for SIMD-intrinsic code generation, comprising 136 carefully crafted tasks and targeting five representative SIMD intrinsics: SSE (x86 Streaming SIMD Extension), AVX (x86 Advanced Vector Extension), Neon (ARM Advanced SIMD Extension), SVE (ARM Scalable Vector Extension), and RVV (RISC-V Vector Extension). We conduct a systematic evaluation (measuring both correctness and performance) of 18 representative LLMs on SimdBench, resulting in a series of novel and insightful findings. Our evaluation results demonstrate that LLMs exhibit a universal decrease in pass@k during SIMD-intrinsic code generation compared to scalar-code generation. Our in-depth analysis highlights promising directions for the further advancement of LLMs in the challenging domain of SIMD-intrinsic code generation. SimdBench is fully open source at https://anonymous.4open.science/r/SimdBench-1B3F/ to benefit the broader research community.

cs.SE↗

arXiv2025

QuantEIT: Ultra-Lightweight Quantum-Assisted Inference for Chest Electrical Impedance Tomography

Electrical Impedance Tomography (EIT) is a non-invasive, low-cost bedside imaging modality with high temporal resolution, making it suitable for bedside monitoring. However, its inherently ill-posed inverse problem poses significant challenges for accurate image reconstruction. Deep learning (DL)-based approaches have shown promise but often rely on complex network architectures with a large number of parameters, limiting efficiency and scalability. Here, we propose an Ultra-Lightweight Quantum-Assisted Inference (QuantEIT) framework for EIT image reconstruction. QuantEIT leverages a Quantum-Assisted Network (QA-Net), combining parallel 2-qubit quantum circuits to generate expressive latent representations that serve as implicit nonlinear priors, followed by a single linear layer for conductivity reconstruction. This design drastically reduces model complexity and parameter number. Uniquely, QuantEIT operates in an unsupervised, training-data-free manner and represents the first integration of quantum circuits into EIT image reconstruction. Extensive experiments on simulated and real-world 2D and 3D EIT lung imaging data demonstrate that QuantEIT outperforms conventional methods, achieving comparable or superior reconstruction accuracy using only 0.2% of the parameters, with enhanced robustness to noise.

cs.CV↗

arXiv2025

D2IP: Deep Dynamic Image Prior for 3D Time-sequence Pulmonary Impedance Imaging

Unsupervised learning methods, such as Deep Image Prior (DIP), have shown great potential in tomographic imaging due to their training-data-free nature and high generalization capability. However, their reliance on numerous network parameter iterations results in high computational costs, limiting their practical application, particularly in complex 3D or time-sequence tomographic imaging tasks. To overcome these challenges, we propose Deep Dynamic Image Prior (D2IP), a novel framework for 3D time-sequence imaging. D2IP introduces three key strategies - Unsupervised Parameter Warm-Start (UPWS), Temporal Parameter Propagation (TPP), and a customized lightweight reconstruction backbone, 3D-FastResUNet - to accelerate convergence, enforce temporal coherence, and improve computational efficiency. Experimental results on both simulated and clinical pulmonary datasets demonstrate that D2IP enables fast and accurate 3D time-sequence Electrical Impedance Tomography (tsEIT) reconstruction. Compared to state-of-the-art baselines, D2IP delivers superior image quality, with a 24.8% increase in average MSSIM and an 8.1% reduction in ERR, alongside significantly reduced computational time (7.1x faster), highlighting its promise for clinical dynamic pulmonary imaging.

eess.IV↗

arXiv2025

AlphaVAE: Unified End-to-End RGBA Image Reconstruction and Generation with Alpha-Aware Representation Learning

Recent advances in latent diffusion models have achieved remarkable results in high-fidelity RGB image synthesis by leveraging pretrained VAEs to compress and reconstruct pixel data at low computational cost. However, the generation of transparent or layered content (RGBA image) remains largely unexplored, due to the lack of large-scale benchmarks. In this work, we propose ALPHA, the first comprehensive RGBA benchmark that adapts standard RGB metrics to four-channel images via alpha blending over canonical backgrounds. We further introduce ALPHAVAE, a unified end-to-end RGBA VAE that extends a pretrained RGB VAE by incorporating a dedicated alpha channel. The model is trained with a composite objective that combines alpha-blended pixel reconstruction, patch-level fidelity, perceptual consistency, and dual KL divergence constraints to ensure latent fidelity across both RGB and alpha representations. Our RGBA VAE, trained on only 8K images in contrast to 1M used by prior methods, achieves a +4.9 dB improvement in PSNR and a +3.2% increase in SSIM over LayerDiffuse in reconstruction. It also enables superior transparent image generation when fine-tuned within a latent diffusion framework. Our code, data, and models are released on https://github.com/o0o0o00o0/AlphaVAE for reproducibility.

cs.CV↗

arXiv2025

Sharp uniform approximation for spectral Barron functions by deep neural networks

This work explores the neural network approximation capabilities for functions within the spectral Barron space $\mathscr{B}^s$, where $s$ is the smoothness index. We demonstrate that for functions in $\mathscr{B}^{1/2}$, a shallow neural network (a single hidden layer) with $N$ units can achieve an $L^p$-approximation rate of $\mathcal{O}(N^{-1/2})$. This rate also applies to uniform approximation, differing by at most a logarithmic factor. Our results significantly reduce the smoothness requirement compared to existing theory, which necessitate functions to belong to $\mathscr{B}^1$ in order to attain the same rate. Furthermore, we show that increasing the network's depth can notably improve the approximation order for functions with small smoothness. Specifically, for networks with $L$ hidden layers, functions in $\mathscr{B}^s$ with $0 < sL \le 1/2$ can achieve an approximation rate of $\mathcal{O}(N^{-sL})$. The rates and prefactors in our estimates are dimension-free. We also confirm the sharpness of our findings, with the lower bound closely aligning with the upper, with a discrepancy of at most one logarithmic factor.

math.NA↗

arXiv2025

Quantization without Tears

Deep neural networks, while achieving remarkable success across diverse tasks, demand significant resources, including computation, GPU memory, bandwidth, storage, and energy. Network quantization, as a standard compression and acceleration technique, reduces storage costs and enables potential inference acceleration by discretizing network weights and activations into a finite set of integer values. However, current quantization methods are often complex and sensitive, requiring extensive task-specific hyperparameters, where even a single misconfiguration can impair model performance, limiting generality across different models and tasks. In this paper, we propose Quantization without Tears (QwT), a method that simultaneously achieves quantization speed, accuracy, simplicity, and generality. The key insight of QwT is to incorporate a lightweight additional structure into the quantized network to mitigate information loss during quantization. This structure consists solely of a small set of linear layers, keeping the method simple and efficient. More importantly, it provides a closed-form solution, allowing us to improve accuracy effortlessly under 2 minutes. Extensive experiments across various vision, language, and multimodal tasks demonstrate that QwT is both highly effective and versatile. In fact, our approach offers a robust solution for network quantization that combines simplicity, accuracy, and adaptability, which provides new insights for the design of novel quantization paradigms. The code is publicly available at https://github.com/wujx2001/QwT

cs.CV↗

arXiv2025

Hamilton Cycles In Vertex-Transitive Graphs of Order 10p

After long-term efforts, the Hamilton path (cycle) problem for connected vertex-transitive graphs of order $pq$ (where $p$ and $q$ are primes) was finally resolved in 2021, see [10]. Fifteen years ago, mathematicians began addressing this problem for graphs of order $2pq$. Among these studies, it was proved in 2012 (see [21]) that every connected vertex-transitive graph of order $10p$ (where $p \neq 7$ is a prime) contains a Hamilton path, with the exception of a family of graphs that was recently confirmed in [11]. In this paper, we achieve a further result: every connected vertex-transitive graph of order $10p$ (where $p$ is a prime) contains a Hamilton cycle, except for the truncation of the Petersen graph.

math.CO↗

arXiv2025

TACOS: Open Tagging and Comparative Scoring for Instruction Fine-Tuning Data Selection

Instruction Fine-Tuning (IFT) is crucial for aligning large language models (LLMs) with human preferences, and selecting a small yet representative subset from massive data significantly facilitates IFT in terms of both efficiency and effectiveness. Nevertheless, existing approaches suffer from two limitations: the use of simple heuristics restricts data diversity, while the singleton data quality evaluation accounts for inconsistent criteria between independent samples. To address the issues, we present TACOS, an innovative method that integrates Open Tagging and Comparative Scoring for IFT data selection. To capture data diversity, we leverage LLMs to assign open-domain tags to human queries, followed by a normalization stage to denoise the open tags and enable efficient clustering. Additionally, we suggest a comparative scoring method that allows the relative quality evaluation of samples within a cluster, avoiding inconsistent criteria seen in singleton-based evaluations. Extensive experiments across diverse datasets and LLM architectures demonstrate that TACOS outperforms existing approaches by a large margin. Notably, it achieves superior instruction-following performance on MT-Bench and ranks 1st among LLaMA2-7B-Based models on AlpacaEval 2.0, illustrating its efficacy for IFT data selection.

cs.CL↗

arXiv2025

Combinatorial p-th Calabi flow for finite and infinite ideal circle patterns

This paper presents a comprehensive study of the combinatorial $p$-th Calabi flow for both finite and infinite ideal circle patterns. In the finite case, we establish a sharp criterion: the combinatorial $p$-th Calabi flow with $p>1$ converges if and only if a constant curvature metric exists in the underlying geometric background. In the infinite setting, we prove the long-time existence of solutions to the combinatorial $p$-th Calabi flow for $p \geq 2$, representing a significant advance in the theory of curvature flows on infinite structures.

math.GT↗

arXiv2025

Characterization of Infinite Ideal Polyhedra in Hyperbolic 3-Space via Combinatorial Ricci Flow

In his seminal work \cite{Ri96}, Rivin characterized finite ideal polyhedra in three-dimensional hyperbolic space. However, the characterization of infinite ideal polyhedra, as proposed by Rivin, has remained a long-standing open problem. In this paper, we introduce the combinatorial Ricci flow for infinite ideal circle patterns, a discrete analogue of Ricci flow on non-compact Riemannian manifolds, and prove a characterization of such circle patterns under certain combinatorial conditions. Our results provide affirmative solutions to Rivin's problem.

math.GT↗