arXiv subjects

Wei Wu

Publications and source records attributed to Wei Wu.

At least 181 records · Page 10Linked to original sources

arXiv2025

Benchmarking Large Vision-Language Models via Directed Scene Graph for Comprehensive Image Captioning

Generating detailed captions comprehending text-rich visual content in images has received growing attention for Large Vision-Language Models (LVLMs). However, few studies have developed benchmarks specifically tailored for detailed captions to measure their accuracy and comprehensiveness. In this paper, we introduce a detailed caption benchmark, termed as CompreCap, to evaluate the visual context from a directed scene graph view. Concretely, we first manually segment the image into semantically meaningful regions (i.e., semantic segmentation mask) according to common-object vocabulary, while also distinguishing attributes of objects within all those regions. Then directional relation labels of these objects are annotated to compose a directed scene graph that can well encode rich compositional information of the image. Based on our directed scene graph, we develop a pipeline to assess the generated detailed captions from LVLMs on multiple levels, including the object-level coverage, the accuracy of attribute descriptions, the score of key relationships, etc. Experimental results on the CompreCap dataset confirm that our evaluation method aligns closely with human evaluation scores across LVLMs.

cs.CV↗

arXiv2025

Learning Visual Generative Priors without Text

Although text-to-image (T2I) models have recently thrived as visual generative priors, their reliance on high-quality text-image pairs makes scaling up expensive. We argue that grasping the cross-modality alignment is not a necessity for a sound visual generative prior, whose focus should be on texture modeling. Such a philosophy inspires us to study image-to-image (I2I) generation, where models can learn from in-the-wild images in a self-supervised manner. We first develop a pure vision-based training framework, Lumos, and confirm the feasibility and the scalability of learning I2I models. We then find that, as an upstream task of T2I, our I2I model serves as a more foundational visual prior and achieves on-par or better performance than existing T2I models using only 1/10 text-image pairs for fine-tuning. We further demonstrate the superiority of I2I priors over T2I priors on some text-irrelevant visual generative tasks, like image-to-3D and image-to-video. Our project page is available at https://ant-research.github.io/lumos.

cs.CV↗

arXiv2025

Non-equilibrium Capacitance Loss in EDLC Based on Ion Flux

Understanding the loss of non-equilibrium capacitance in the electric double layer capacitor (EDLC) during continuous charging compared to equilibrium capacitance is crucial for the design of high-rate energy storage devices. Due to three major challenges, including strong ion correlations induced by high concentrations at the interface, the difficulty in representing far from equilibrium, and numerical stability issues, current numerical simulations of ion distribution based on statistical mechanics struggle to achieve long-term accurate modeling under realistic parameters. Since non-equilibrium capacitance loss is a universal macroscopic phenomenon in EDLC, which exists independently of the microscopic model of the electrolyte, this study starts from macroscopic non-equilibrium phenomena, introduces a novel perspective centered on ion flux (IF) for the first time. By deriving the relationship among the charging current, IF, and displacement current, a novel capacitance model based on IF is first established. It is further demonstrated that the current relationship and voltage relationship of EDLC, when represented by conventional circuit models, exhibit inherent contradictions. Therefore, a new interpretative framework is proposed, wherein IF and displacement current are analogized to acceleration and deformation in an elastic medium, respectively. Inspired by this insight, this study reveals that the decrease in non-equilibrium capacitance with increasing charging current originates from the inherent resistance of IF to external charging. This tendency to restore equilibrium is the fundamental cause of non-equilibrium capacitance loss.

cond-mat.mtrl-sci↗

arXiv2025

Cross-Layer Security for Semantic Communications: Metrics and Optimization

Different from traditional secure communication that focuses on symbolic protection at the physical layer, semantic secure communication requires further attention to semantic-level task performance at the application layer. There is a research gap on how to comprehensively evaluate and optimize the security performance of semantic communication. In order to fill this gap, a unified semantic security metric, the cross-layer semantic secure rate (CLSSR), is defined to estimate cross-layer security requirements at both the physical layer and the application layer. Then, we formulate the maximization problem of the CLSSR with the mixed integer nonlinear programming (MINLP). We propose a hierarchical AI-native semantic secure communication network with a reinforcement learning (RL)-based semantic resource allocation scheme, aiming to ensure the cross-layer semantic security (CL-SS). Finally, we prove the convergence of our proposed intelligent resource allocation, and the simulation results demonstrate that our proposed CLSS method outperforms the traditional physical layer semantic security (PL-SS) method in terms of both task reliability and CLSSR.

cs.IT↗

arXiv2025

Theoretical Investigation of High-Tc Superconductivity in Sr-Doped La$_3$Ni$_2$O$_7$ at Ambient Pressure

The recent discovery of pressure-induced superconductivity in La$_3$Ni$_2$O$_7$ has established a novel platform for studying unconventional superconductors. However, achieving superconductivity in this system currently requires relatively high pressures. In this study, we propose a chemical pressure strategy via Sr substitution to stabilize high-Tc superconductivity in La$_3$Ni$_2$O$_7$ under ambient conditions.Using density functional theory (DFT) calculations, we systematically investigate the structural and electronic properties of Sr-doped La$_{3-x}$Sr$_x$Ni$_2$O$_7$ (x = 0.25, 0.5, 1) at ambient pressure and identify two dynamically stable phases:La$_{2.5}$Sr$_{0.5}$Ni$_2$O$_7$ and La$_2$SrNi$_2$O$_7$.Our calculations reveal that both phases exhibit metallization of the $σ$-bonding bands dominated by Ni-d$_{z^{2}}$ orbitals-a key feature associated with high-Tc superconductivity, as reported in the high-pressure phase of La$_3$Ni$_2$O$_7$. Further analysis using tight-binding models shows that the key hopping parameters in La$_{2.5}$Sr$_{0.5}$Ni$_2$O$_7$ and La$_2$SrNi$_2$O$_7$ closely resemble those of La$_3$Ni$_2$O$_7$ under high pressure, indicating that strong super-exchange interactions between interlayer Ni-d$_{z^{2}}$ orbitals are preserved. These findings suggest that Sr-doped La$_3$Ni$_2$O$_7$ is a promising candidate for realizing high-Tcsuperconductivity at ambient pressure.

cond-mat.supr-con↗

arXiv2025

Hierarchical Locality Sensitive Hashing for Structured Data: A Survey

Data similarity (or distance) computation is a fundamental research topic which fosters a variety of similarity-based machine learning and data mining applications. In big data analytics, it is impractical to compute the exact similarity of data instances due to high computational cost. To this end, the Locality Sensitive Hashing (LSH) technique has been proposed to provide accurate estimators for various similarity measures between sets or vectors in an efficient manner without the learning process. Structured data (e.g., sequences, trees and graphs), which are composed of elements and relations between the elements, are commonly seen in the real world, but the traditional LSH algorithms cannot preserve the structure information represented as relations between elements. In order to conquer the issue, researchers have been devoted to the family of the hierarchical LSH algorithms. In this paper, we explore the present progress of the research into hierarchical LSH from the following perspectives: 1) Data structures, where we review various hierarchical LSH algorithms for three typical data structures and uncover their inherent connections; 2) Applications, where we review the hierarchical LSH algorithms in multiple application scenarios; 3) Challenges, where we discuss some potential challenges as future directions.

cs.DS↗

arXiv2025

Every FLOP Counts: Scaling a 300B Mixture-of-Experts LING LLM without Premium GPUs

In this technical report, we tackle the challenges of training large-scale Mixture of Experts (MoE) models, focusing on overcoming cost inefficiency and resource limitations prevalent in such systems. To address these issues, we present two differently sized MoE large language models (LLMs), namely Ling-Lite and Ling-Plus (referred to as "Bailing" in Chinese, spelled Bǎilíng in Pinyin). Ling-Lite contains 16.8 billion parameters with 2.75 billion activated parameters, while Ling-Plus boasts 290 billion parameters with 28.8 billion activated parameters. Both models exhibit comparable performance to leading industry benchmarks. This report offers actionable insights to improve the efficiency and accessibility of AI development in resource-constrained settings, promoting more scalable and sustainable technologies. Specifically, to reduce training costs for large-scale MoE models, we propose innovative methods for (1) optimization of model architecture and training processes, (2) refinement of training anomaly handling, and (3) enhancement of model evaluation efficiency. Additionally, leveraging high-quality data generated from knowledge graphs, our models demonstrate superior capabilities in tool use compared to other models. Ultimately, our experimental findings demonstrate that a 300B MoE LLM can be effectively trained on lower-performance devices while achieving comparable performance to models of a similar scale, including dense and MoE models. Compared to high-performance devices, utilizing a lower-specification hardware system during the pre-training phase demonstrates significant cost savings, reducing computing costs by approximately 20%. The models can be accessed at https://huggingface.co/inclusionAI.

cs.LG↗

arXiv2025

Exploring atom-ion Feshbach resonances below the s-wave limit

Revealing the quantum properties of matter requires a high degree of experimental control accompanied by a profound theoretical understanding. At ultracold temperatures, quantities that appear continuous in everyday life, such as the motional angular momentum of two colliding particles, become quantized, leaving a measurable imprint on experimental results. Embedding a single particle within a larger quantum bath at lowest temperatures can result in resonant partial-wave dependent interaction, whose strength near zero energy is dictated by universal threshold scaling laws. Hybrid atom-ion systems have emerged as a novel platform in which a single charged atom in an ultracold bath serves as a well-controlled impurity of variable energy. However, entering the low-energy s-wave regime and exploring the role of higher-partial-wave scattering within has remained an open challenge. Here, we immerse a Barium ion in a cloud of ultracold spin-polarized Lithium atoms, realize tunable collision energies below the s-wave limit and explore resonant higher-partial-wave scattering by studying the energy dependence of Feshbach resonances. Utilizing precise electric field control, we tune the collision energy over four orders of magnitude, reaching from the many-parital-wave to the s-wave regime. At the lowest energies, we probe the energy dependence of an isolated s-wave Feshbach resonance and introduce a theoretical model that allows to distinguish it from higher-partial-wave resonances. Additionally, at energies around the p-wave barrier, we find and identify an open-channel f-wave resonance, consistent with threshold laws. Our findings highlight and benchmark the importance of higher-partial-wave scattering well within the s-wave regime and offer control over chemical reactions and complex many-body dynamics in atom-ion ensembles - on the level of individual angular momentum quanta.

physics.atom-ph↗

arXiv2025

Exploring the Performance Improvement of Tensor Processing Engines through Transformation in the Bit-weight Dimension of MACs

General matrix-matrix multiplication (GEMM) is a cornerstone of AI computations, making tensor processing engines (TPEs) increasingly critical in GPUs and domain-specific architectures. Existing architectures primarily optimize dataflow or operand reuse strategies. However, considering the interaction between matrix multiplication and multiply-accumulators (MACs) offers greater optimization potential. This work introduces a novel hardware perspective on matrix multiplication, focusing on the bit-weight dimension of MACs. We propose a finer-grained TPE notation using matrix triple loops as an example, introducing new methods for designing and optimizing PE microarchitectures. Based on this notation and its transformations, we propose four optimization techniques that improve timing, area, and power consumption. Implementing our design in RTL using the SMIC-28nm process, we evaluate its effectiveness across four classic TPE architectures: systolic array, 3D-Cube, multiplier-adder tree, and 2D-Matrix. Our techniques achieve area efficiency improvements of 1.27x, 1.28x, 1.56x, and 1.44x, and energy efficiency gains of 1.04x, 1.56x, 1.49x, and 1.20x, respectively. Applied to a bit-slice architecture, our approach achieves a 12.10x improvement in energy efficiency and 2.85x in area efficiency compared to Laconic. Our Verilog HDL code, along with timing, area, and power reports, is available at https://github.com/wqzustc/High-Performance-Tensor-Processing-Engines

cs.AR↗

arXiv2025

RoboSense: Large-scale Dataset and Benchmark for Egocentric Robot Perception and Navigation in Crowded and Unstructured Environments

Reliable embodied perception from an egocentric perspective is challenging yet essential for autonomous navigation technology of intelligent mobile agents. With the growing demand of social robotics, near-field scene understanding becomes an important research topic in the areas of egocentric perceptual tasks related to navigation in both crowded and unstructured environments. Due to the complexity of environmental conditions and difficulty of surrounding obstacles owing to truncation and occlusion, the perception capability under this circumstance is still inferior. To further enhance the intelligence of mobile robots, in this paper, we setup an egocentric multi-sensor data collection platform based on 3 main types of sensors (Camera, LiDAR and Fisheye), which supports flexible sensor configurations to enable dynamic sight of view from ego-perspective, capturing either near or farther areas. Meanwhile, a large-scale multimodal dataset is constructed, named RoboSense, to facilitate egocentric robot perception. Specifically, RoboSense contains more than 133K synchronized data with 1.4M 3D bounding box and IDs annotated in the full $360^{\circ}$ view, forming 216K trajectories across 7.6K temporal sequences. It has $270\times$ and $18\times$ as many annotations of surrounding obstacles within near ranges as the previous datasets collected for autonomous driving scenarios such as KITTI and nuScenes. Moreover, we define a novel matching criterion for near-field 3D perception and prediction metrics. Based on RoboSense, we formulate 6 popular tasks to facilitate the future research development, where the detailed analysis as well as benchmarks are also provided accordingly. Data desensitization measures have been conducted for privacy protection.

cs.CV↗

arXiv2025

PromptCoT: Synthesizing Olympiad-level Problems for Mathematical Reasoning in Large Language Models

The ability of large language models to solve complex mathematical problems has progressed significantly, particularly for tasks requiring advanced reasoning. However, the scarcity of sufficiently challenging problems, particularly at the Olympiad level, hinders further advancements. In this work, we introduce PromptCoT, a novel approach for automatically generating high-quality Olympiad-level math problems. The proposed method synthesizes complex problems based on mathematical concepts and the rationale behind problem construction, emulating the thought processes of experienced problem designers. We provide a theoretical analysis demonstrating that an optimal rationale should maximize both the likelihood of rationale generation given the associated concepts and the likelihood of problem generation conditioned on both the rationale and the concepts. Our method is evaluated on standard benchmarks including GSM8K, MATH-500, and AIME2024, where it consistently outperforms existing problem generation methods. Furthermore, we demonstrate that PromptCoT exhibits superior data scalability, consistently maintaining high performance as the dataset size increases, outperforming the baselines. The implementation is available at https://github.com/zhaoxlpku/PromptCoT.

cs.CL↗

arXiv2025

Neuron Platonic Intrinsic Representation From Dynamics Using Contrastive Learning

The Platonic Representation Hypothesis suggests a universal, modality-independent reality representation behind different data modalities. Inspired by this, we view each neuron as a system and detect its multi-segment activity data under various peripheral conditions. We assume there's a time-invariant representation for the same neuron, reflecting its intrinsic properties like molecular profiles, location, and morphology. The goal of obtaining these intrinsic neuronal representations has two criteria: (I) segments from the same neuron should have more similar representations than those from different neurons; (II) the representations must generalize well to out-of-domain data. To meet these, we propose the NeurPIR (Neuron Platonic Intrinsic Representation) framework. It uses contrastive learning, with segments from the same neuron as positive pairs and those from different neurons as negative pairs. In implementation, we use VICReg, which focuses on positive pairs and separates dissimilar samples via regularization. We tested our method on Izhikevich model-simulated neuronal population dynamics data. The results accurately identified neuron types based on preset hyperparameters. We also applied it to two real-world neuron dynamics datasets with neuron type annotations from spatial transcriptomics and neuron locations. Our model's learned representations accurately predicted neuron types and locations and were robust on out-of-domain data (from unseen animals). This shows the potential of our approach for understanding neuronal systems and future neuroscience research.

q-bio.NC↗

arXiv2025

Classifying the Stoichiometry of Virus-like Particles with Interpretable Machine Learning

Virus-like particles (VLPs) are valuable for vaccine development due to their immune-triggering properties. Understanding their stoichiometry, the number of protein subunits to form a VLP, is critical for vaccine optimisation. However, current experimental methods to determine stoichiometry are time-consuming and require highly purified proteins. To efficiently classify stoichiometry classes in proteins, we curate a new dataset and propose an interpretable, data-driven pipeline leveraging linear machine learning models. We also explore the impact of feature encoding on model performance and interpretability, as well as methods to identify key protein sequence features influencing classification. The evaluation of our pipeline demonstrates that it can classify stoichiometry while revealing protein features that possibly influence VLP assembly. The data and code used in this work are publicly available at https://github.com/Shef-AIRE/StoicIML.

cs.LG↗

arXiv2025

Observation of quantum temporal correlations well beyond Luders bound

Quantum correlations, both spatial and temporal, are the central pillars of quantum mechanics. Over the last two decades, a big breakthrough in quantum physics is its complex extension to the non-Hermitian realm, and dizzying varieties of novel phenomena and applications beyond the Hermitian framework have been uncovered. However, unique features of non-Hermitian quantum correlations, especially in the time domain, still remain to be explored. Here, for the first time, we experimentally achieve this goal by using a parity-time (PT)-symmetric trapped-ion system. The upper limit of temporal quantum correlations, known as the algebraic bound, which has so far not been achieved in the standard measurement scenario, is reached here by approaching the exceptional point (EP), thus showing the unexpected ability of EPs in tuning temporal quantum correlation effects. Our study, unveiling the fundamental interplay of non-Hermiticity, nonlinearity, and temporal quantum correlations, provides the first step towards exploring and utilizing various non-Hermitian temporal quantum effects by operating a wide range of EP devices, which are important for both fundamental studies and applications of quantum EP systems.

quant-ph↗

arXiv2025

LCSim: A Large-Scale Controllable Traffic Simulator

With the rapid growth of urban transportation and the continuous progress in autonomous driving, a demand for robust benchmarking autonomous driving algorithms has emerged, calling for accurate modeling of large-scale urban traffic scenarios with diverse vehicle driving styles. Traditional traffic simulators, such as SUMO, often depend on hand-crafted scenarios and rule-based models, where vehicle actions are limited to speed adjustment and lane changes, making it difficult for them to create realistic traffic environments. In recent years, real-world traffic scenario datasets have been developed alongside advancements in autonomous driving, facilitating the rise of data-driven simulators and learning-based simulation methods. However, current data-driven simulators are often restricted to replicating the traffic scenarios and driving styles within the datasets they rely on, limiting their ability to model multi-style driving behaviors observed in the real world. We propose \textit{LCSim}, a large-scale controllable traffic simulator. First, we define a unified data format for traffic scenarios and provide tools to construct them from multiple data sources, enabling large-scale traffic simulation. Furthermore, we integrate a diffusion-based vehicle motion planner into LCSim to facilitate realistic and diverse vehicle modeling. Under specific guidance, this allows for the creation of traffic scenarios that reflect various driving styles. Leveraging these features, LCSim can provide large-scale, realistic, and controllable virtual traffic environments. Codes and demos are available at https://tsinghua-fib-lab.github.io/LCSim.

cs.RO↗

arXiv2025

BLEND: Behavior-guided Neural Population Dynamics Modeling via Privileged Knowledge Distillation

Modeling the nonlinear dynamics of neuronal populations represents a key pursuit in computational neuroscience. Recent research has increasingly focused on jointly modeling neural activity and behavior to unravel their interconnections. Despite significant efforts, these approaches often necessitate either intricate model designs or oversimplified assumptions. Given the frequent absence of perfectly paired neural-behavioral datasets in real-world scenarios when deploying these models, a critical yet understudied research question emerges: how to develop a model that performs well using only neural activity as input at inference, while benefiting from the insights gained from behavioral signals during training? To this end, we propose BLEND, the behavior-guided neural population dynamics modeling framework via privileged knowledge distillation. By considering behavior as privileged information, we train a teacher model that takes both behavior observations (privileged features) and neural activities (regular features) as inputs. A student model is then distilled using only neural activity. Unlike existing methods, our framework is model-agnostic and avoids making strong assumptions about the relationship between behavior and neural activity. This allows BLEND to enhance existing neural dynamics modeling architectures without developing specialized models from scratch. Extensive experiments across neural population activity modeling and transcriptomic neuron identity prediction tasks demonstrate strong capabilities of BLEND, reporting over 50% improvement in behavioral decoding and over 15% improvement in transcriptomic neuron identity prediction after behavior-guided distillation. Furthermore, we empirically explore various behavior-guided distillation strategies within the BLEND framework and present a comprehensive analysis of effectiveness and implications for model performance.

cs.NE↗

arXiv2025

Car-GS: Addressing Reflective and Transparent Surface Challenges in 3D Car Reconstruction

3D car modeling is crucial for applications in autonomous driving systems, virtual and augmented reality, and gaming. However, due to the distinctive properties of cars, such as highly reflective and transparent surface materials, existing methods often struggle to achieve accurate 3D car reconstruction.To address these limitations, we propose Car-GS, a novel approach designed to mitigate the effects of specular highlights and the coupling of RGB and geometry in 3D geometric and shading reconstruction (3DGS). Our method incorporates three key innovations: First, we introduce view-dependent Gaussian primitives to effectively model surface reflections. Second, we identify the limitations of using a shared opacity parameter for both image rendering and geometric attributes when modeling transparent objects. To overcome this, we assign a learnable geometry-specific opacity to each 2D Gaussian primitive, dedicated solely to rendering depth and normals. Third, we observe that reconstruction errors are most prominent when the camera view is nearly orthogonal to glass surfaces. To address this issue, we develop a quality-aware supervision module that adaptively leverages normal priors from a pre-trained large-scale normal model.Experimental results demonstrate that Car-GS achieves precise reconstruction of car surfaces and significantly outperforms prior methods. The project page is available at https://lcc815.github.io/Car-GS.

cs.CV↗

arXiv2025

LLMs as Workers in Human-Computational Algorithms? Replicating Crowdsourcing Pipelines with LLMs

LLMs have shown promise in replicating human-like behavior in crowdsourcing tasks that were previously thought to be exclusive to human abilities. However, current efforts focus mainly on simple atomic tasks. We explore whether LLMs can replicate more complex crowdsourcing pipelines. We find that modern LLMs can simulate some of crowdworkers' abilities in these ``human computation algorithms,'' but the level of success is variable and influenced by requesters' understanding of LLM capabilities, the specific skills required for sub-tasks, and the optimal interaction modality for performing these sub-tasks. We reflect on human and LLMs' different sensitivities to instructions, stress the importance of enabling human-facing safeguards for LLMs, and discuss the potential of training humans and LLMs with complementary skill sets. Crucially, we show that replicating crowdsourcing pipelines offers a valuable platform to investigate 1) the relative LLM strengths on different tasks (by cross-comparing their performances on sub-tasks) and 2) LLMs' potential in complex tasks, where they can complete part of the tasks while leaving others to humans.

cs.CL↗