arXiv subjects

Peng Wang

Publications and source records attributed to Peng Wang.

At least 235 records · Page 13Linked to original sources

arXiv2025

Exploring Efficiency Frontiers of Thinking Budget in Medical Reasoning: Scaling Laws between Computational Resources and Reasoning Quality

This study presents the first comprehensive evaluation of thinking budget mechanisms in medical reasoning tasks, revealing fundamental scaling laws between computational resources and reasoning quality. We systematically evaluated two major model families, Qwen3 (1.7B to 235B parameters) and DeepSeek-R1 (1.5B to 70B parameters), across 15 medical datasets spanning diverse specialties and difficulty levels. Through controlled experiments with thinking budgets ranging from zero to unlimited tokens, we establish logarithmic scaling relationships where accuracy improvements follow a predictable pattern with both thinking budget and model size. Our findings identify three distinct efficiency regimes: high-efficiency (0 to 256 tokens) suitable for real-time applications, balanced (256 to 512 tokens) offering optimal cost-performance tradeoffs for routine clinical support, and high-accuracy (above 512 tokens) justified only for critical diagnostic tasks. Notably, smaller models demonstrate disproportionately larger benefits from extended thinking, with 15 to 20% improvements compared to 5 to 10% for larger models, suggesting a complementary relationship where thinking budget provides greater relative benefits for capacity-constrained models. Domain-specific patterns emerge clearly, with neurology and gastroenterology requiring significantly deeper reasoning processes than cardiovascular or respiratory medicine. The consistency between Qwen3 native thinking budget API and our proposed truncation method for DeepSeek-R1 validates the generalizability of thinking budget concepts across architectures. These results establish thinking budget control as a critical mechanism for optimizing medical AI systems, enabling dynamic resource allocation aligned with clinical needs while maintaining the transparency essential for healthcare deployment.

cs.CL↗

arXiv2025

Seedream 4.0: Toward Next-generation Multimodal Image Generation

We introduce Seedream 4.0, an efficient and high-performance multimodal image generation system that unifies text-to-image (T2I) synthesis, image editing, and multi-image composition within a single framework. We develop a highly efficient diffusion transformer with a powerful VAE which also can reduce the number of image tokens considerably. This allows for efficient training of our model, and enables it to fast generate native high-resolution images (e.g., 1K-4K). Seedream 4.0 is pretrained on billions of text-image pairs spanning diverse taxonomies and knowledge-centric concepts. Comprehensive data collection across hundreds of vertical scenarios, coupled with optimized strategies, ensures stable and large-scale training, with strong generalization. By incorporating a carefully fine-tuned VLM model, we perform multi-modal post-training for training both T2I and image editing tasks jointly. For inference acceleration, we integrate adversarial distillation, distribution matching, and quantization, as well as speculative decoding. It achieves an inference time of up to 1.8 seconds for generating a 2K image (without a LLM/VLM as PE model). Comprehensive evaluations reveal that Seedream 4.0 can achieve state-of-the-art results on both T2I and multimodal image editing. In particular, it demonstrates exceptional multimodal capabilities in complex tasks, including precise image editing and in-context reasoning, and also allows for multi-image reference, and can generate multiple output images. This extends traditional T2I systems into an more interactive and multidimensional creative tool, pushing the boundary of generative AI for both creativity and professional applications. We further scale our model and data as Seedream 4.5. Seedream 4.0 and Seedream 4.5 are accessible on Volcano Engine https://www.volcengine.com/experience/ark?launch=seedream.

cs.CV↗

arXiv2025

Self-Propelled Agents and Group Social Force

Brownian motion have long been studied on a diversity of fields, not only in physics of statistical mechanics, but also in biological models, finance and economic process, and social systems. In the past twenty years, there has been a growing interest in studying the model in self-propelled feature and interaction force such that the model also fits into study of social phenomenon of many individuals. This article will continue with this research trend and especially investigate the model in paradigms for a quantitative description of social and economic process. We mainly discuss a class of collective decision process of Brownian agent/particles, where the stochastic process does not exist in the fluctuation in the traditional Brownian motion, but in selection among several discrete choices. Their decisions interacts with each other in a given social topology. To simplify our discussion the binary choice problem is particularly discussed where each agent only takes an alternative of two choices. Mathematically, we introduce a set of arrays to describe social relationship of agents in a quantitative manner, and the arrays deduce the group social force and opinion dynamics, which are useful to study complex social movement and self-organization phenomena including discrete-choice activities, social groups and de-individualization effect. Such agent-based simulation symbolizes a variety of collective activities in human society, especially in the field of economics and social science.

physics.soc-ph↗

arXiv2025

The Willmore problem for surfaces with symmetry

The Willmore Problem seeks closed surfaces in $\mathbb{S}^3\subset\mathbb{R}^4$ of a given topological type minimizing the squared-mean-curvature energy $W = \int |H_{\mathbb{R}^4}|^2 = area + \int |H_{\mathbb{S}^3}|^2$. The longstanding Willmore Conjecture that the Clifford torus minimizes $W$ among genus-$1$ surfaces is now a theorem of Marques and Neves [22], but the general conjecture [12] that Lawson's [18] minimal surface $ξ_{g,1}\subset\mathbb{S}^3$ minimizes $W$ among surfaces of genus $g>1$ remains open. Here we prove this conjecture under the additional assumption that the competitor surfaces $M\subset\mathbb{S}^3$ share the ambient symmetries $\widehat{G}_{g,1}$ of $ξ_{g,1}$. In fact, we show each Lawson surface $ξ_{m,k}$ satisfies the corresponding $W$-minimizing property under a smaller symmetry group $\widetilde{G}_{m,k}=\widehat{G}_{m,k}\cap SO(4)$. We also describe a genus 2 example where known methods do not ensure the existence of a $W$-minimizer among surfaces with its symmetry.

math.DG↗

arXiv2025

Classification of Willmore $2$-spheres in $S^n$

This paper resolves a long-standing open problem by providing a classification of Willmore $2$-spheres in $S^n$. We show that any such $2$-sphere is either totally isotropic--originating from the projection of a special twistor curve in the twistor bundle over an even-dimensional sphere--or strictly $k$-isotropic, obtained via $(m-k)$ steps of adjoint transforms of a strictly $m$-isotropic minimal surface in $\mathbb{R}^n$, where $0\leq k\leq m\leq[n/2]-2$. Our approach hinges on the construction of a harmonic sequence in the real Grassmannian over the Lorentz space, derived from the harmonic conformal Gauss map of the original Willmore sphere. This sequence terminates finitely and generalizes, in part, the classical theory of harmonic sequences for harmonic $2$-spheres in complex Grassmannians.

math.DG↗

arXiv2025

Quark-Lepton Unification Signatures

We investigate the collider signatures of the minimal framework for quark-lepton unification at a scale not far from the electroweak symmetry breaking scale. This theory predicts a rich spectrum of new fields, including one vector leptoquark, two scalar leptoquarks, a color-octet scalar, and an additional Higgs doublet. Neutrino masses are generated via the inverse seesaw mechanism, facilitating viable matter unification at the low scale. We find that this theory predicts that in many cases the dominant leptoquark decays are to the third generation Standard Model quarks and leptons. We identify key experimental signatures at the Large Hadron Collider, evaluate and discuss the limits from current measurements, and outline potential strategies for probing this theory in the near future.

hep-ph↗

arXiv2025

Finite- and Large- Sample Inference for Model and Coefficients in High-dimensional Linear Regression with Repro Samples

In this paper, we present a novel and effective inference approach to conduct both finite- and large-sample inference for high-dimensional linear regression models. This approach is developed under the so-called repro samples framework, in which we conduct statistical inference by creating and studying the behavior of artificial samples that are obtained by mimicking the sampling mechanism of the data. We construct confidence sets for (a) the true model corresponding to the nonzero coefficients, (b) a single or any collection of regression coefficients, and (c) both the model and regression coefficients jointly. To facilitate the constructions of these confidence sets and overcome computational difficulties of searching all possible models, we use an innovative Fisher inversion technique to construct a model candidate set that includes the true sparse model with the probability close to 1 for models with both Gaussian and non-Gaussian errors. The proposed approach fills in two major gaps in the high-dimensional regression literature: (1) lack of effective approaches to addressing model selection uncertainty and providing valid inference for the underlying true model; (2) lack of effective inference approaches to guaranteeing finite-sample performance. We provide both finite-sample and asymptotic results to theoretically guarantee the performance of the proposed methods. In addition, our numerical results demonstrate that the proposed methods are valid and achieve better coverage with smaller confidence sets than the current state-of-the-art approaches, such as debiasing and bootstrap approaches.

stat.ME↗

arXiv2025

Qwen3-VL Technical Report

We introduce Qwen3-VL, the most capable vision-language model in the Qwen series to date, achieving superior performance across a broad range of multimodal benchmarks. It natively supports interleaved contexts of up to 256K tokens, seamlessly integrating text, images, and video. The model family includes both dense (2B/4B/8B/32B) and mixture-of-experts (30B-A3B/235B-A22B) variants to accommodate diverse latency-quality trade-offs. Qwen3-VL delivers three core pillars: (i) markedly stronger pure-text understanding, surpassing comparable text-only backbones in several cases; (ii) robust long-context comprehension with a native 256K-token window for both text and interleaved multimodal inputs, enabling faithful retention, retrieval, and cross-referencing across long documents and videos; and (iii) advanced multimodal reasoning across single-image, multi-image, and video tasks, demonstrating leading performance on comprehensive evaluations such as MMMU and visual-math benchmarks (e.g., MathVista and MathVision). Architecturally, we introduce three key upgrades: (i) an enhanced interleaved-MRoPE for stronger spatial-temporal modeling across images and video; (ii) DeepStack integration, which effectively leverages multi-level ViT features to tighten vision-language alignment; and (iii) text-based time alignment for video, evolving from T-RoPE to explicit textual timestamp alignment for more precise temporal grounding. Under comparable token budgets and latency constraints, Qwen3-VL achieves superior performance in both dense and Mixture-of-Experts (MoE) architectures. We envision Qwen3-VL serving as a foundational engine for image-grounded reasoning, agentic decision-making, and multimodal code intelligence in real-world workflows.

cs.CV↗

arXiv2025

SFA: Scan, Focus, and Amplify toward Guidance-aware Answering for Video TextVQA

Video text-based visual question answering (Video TextVQA) task aims to answer questions about videos by leveraging the visual text appearing within the videos. This task poses significant challenges, requiring models to accurately perceive and comprehend scene text that varies in scale, orientation, and clarity across frames, while effectively integrating temporal and semantic context to generate precise answers. Moreover, the model must identify question-relevant textual cues and filter out redundant or irrelevant information to ensure answering is guided by the most relevant and informative cues. To address these challenges, we propose SFA, a training-free framework and the first Video-LLM-based method tailored for Video TextVQA, motivated by the human process of answering questions. By adaptively scanning video frames, selectively focusing on key regions, and directly amplifying them, SFA effectively guides the Video-LLM's attention toward essential cues, enabling it to generate more accurate answers. SFA achieves new state-of-the-art results across several public Video TextVQA datasets and surpasses previous methods by a substantial margin, demonstrating its effectiveness and generalizability.

cs.CV↗

arXiv2025

DAS: Dual-Aligned Semantic IDs Empowered Industrial Recommender System

Semantic IDs are discrete identifiers generated by quantizing the Multi-modal Large Language Models (MLLMs) embeddings, enabling efficient multi-modal content integration in recommendation systems. However, their lack of collaborative signals results in a misalignment with downstream discriminative and generative recommendation objectives. Recent studies have introduced various alignment mechanisms to address this problem, but their two-stage framework design still leads to two main limitations: (1) inevitable information loss during alignment, and (2) inflexibility in applying adaptive alignment strategies, consequently constraining the mutual information maximization during the alignment process. To address these limitations, we propose a novel and flexible one-stage Dual-Aligned Semantic IDs (DAS) method that simultaneously optimizes quantization and alignment, preserving semantic integrity and alignment quality while avoiding the information loss typically associated with two-stage methods. Meanwhile, DAS achieves more efficient alignment between the semantic IDs and collaborative signals, with the following two innovative and effective approaches: (1) Multi-view Constrative Alignment: To maximize mutual information between semantic IDs and collaborative signals, we first incorporate an ID-based CF debias module, and then design three effective contrastive alignment methods: dual user-to-item (u2i), dual item-to-item/user-to-user (i2i/u2u), and dual co-occurrence item-to-item/user-to-user (i2i/u2u). (2) Dual Learning: By aligning the dual quantizations of users and ads, the constructed semantic IDs for users and ads achieve stronger alignment. Finally, we conduct extensive offline experiments and online A/B tests to evaluate DAS's effectiveness, which is now successfully deployed across various advertising scenarios at Kuaishou App, serving over 400 million users daily.

cs.IR↗

arXiv2025

The Cosmic Dance: Observational Detection of Coherent Spin in Galaxy Clusters

The spin of galaxy clusters encodes key information about their formation, dynamics, and the influence of large-scale structure. However, whether clusters possess statistically significant spin and how to measure it observationally remain open questions. Here, we present the first observational, statistical detection of coherent spin in galaxy clusters, using two samples of 2,170 and 1,329 systems with $M > 10^{14}\,M_\odot$, selected from two publicly available group catalogs (\citet{2017A&A...602A.100T} and \citet{2012ApJ...752...41Y}) constructed with two different algorithms and but both based primarily on SDSS galaxies. Cluster spin is quantified by identifying the orientation in the projected plane that maximizes the redshift difference ($ΔZ_{\rm max}$) between member galaxies in two regions divided by a trial axis. We find compelling statistical evidence for coherent rotation, as the observed $ΔZ_{\rm max}$ distribution departs markedly from the randomized controls, exhibiting pronounced deviations near $380\,\mathrm{km\,s^{-1}}$. Stacked visualizations confirm the spatial segregation of redshifted and blueshifted galaxies across the rotation axis. The radial profile of the rotational velocity indicates that it increases as a function of radius. The cluster rotation speed increases with mass, from $\sim360~\mathrm{km\,s}^{-1}$ at $10^{14} M_\odot$ to $\sim693~\mathrm{km\,s}^{-1}$ at $10^{15} M_\odot$. Additionally, cluster spin tends to align parallel with the central galaxy spin and perpendicular to the nearest cosmic filament, particularly in richer systems. These results reveal significant coherent spin in galaxy clusters, shaped by both internal dynamics and large-scale structure.

astro-ph.GA↗

arXiv2025

Fairness in Multi-modal Medical Diagnosis with Demonstration Selection

Multimodal large language models (MLLMs) have shown strong potential for medical image reasoning, yet fairness across demographic groups remains a major concern. Existing debiasing methods often rely on large labeled datasets or fine-tuning, which are impractical for foundation-scale models. We explore In-Context Learning (ICL) as a lightweight, tuning-free alternative for improving fairness. Through systematic analysis, we find that conventional demonstration selection (DS) strategies fail to ensure fairness due to demographic imbalance in selected exemplars. To address this, we propose Fairness-Aware Demonstration Selection (FADS), which builds demographically balanced and semantically relevant demonstrations via clustering-based sampling. Experiments on multiple medical imaging benchmarks show that FADS consistently reduces gender-, race-, and ethnicity-related disparities while maintaining strong accuracy, offering an efficient and scalable path toward fair medical image reasoning. These results highlight the potential of fairness-aware in-context learning as a scalable and data-efficient solution for equitable medical image reasoning.

cs.CV↗

arXiv2025

A Spatial Semantics and Continuity Perception Attention for Remote Sensing Water Body Change Detection

Remote sensing Water Body Change Detection (WBCD) aims to detect water body surface changes from bi-temporal images of the same geographic area. Recently, the scarcity of high spatial resolution datasets for WBCD restricts its application in urban and rural regions, which require more accurate positioning. Meanwhile, previous deep learning-based methods fail to comprehensively exploit the spatial semantic and structural information in deep features in the change detection networks. To resolve these concerns, we first propose a new dataset, HSRW-CD, with a spatial resolution higher than 3 meters for WBCD. Specifically, it contains a large number of image pairs, widely covering various water body types. Besides, a Spatial Semantics and Continuity Perception (SSCP) attention module is designed to fully leverage both the spatial semantics and structure of deep features in the WBCD networks, significantly improving the discrimination capability for water body. The proposed SSCP has three components: the Multi-Semantic spatial Attention (MSA), the Structural Relation-aware Global Attention (SRGA), and the Channel-wise Self-Attention (CSA). The MSA enhances the spatial semantics of water body features and provides precise spatial semantic priors for the CSA. Then, the SRGA further extracts spatial structure to learn the spatial continuity of the water body. Finally, the CSA utilizes the spatial semantic and structural priors from the MSA and SRGA to compute the similarity across channels. Specifically designed as a plug-and-play module for water body deep features, the proposed SSCP allows integration into existing WBCD models. Numerous experiments conducted on the proposed HSRW-CD and Water-CD datasets validate the effectiveness and generalization of the SSCP. The code of this work and the HSRW-CD dataset will be accessed at https://github.com/QingMa1/SSCP.

cs.CV↗

arXiv2025

Halo spin and orientation in Interacting Dark Matter Dark Energy Cosmology

In recent years, the interaction between dark matter (DM) and dark energy (DE) has become a topic of interest in cosmology. Interacting dark matter-dark energy (IDE) models have a substantial impact on the formation of cosmological large-scale structures, which serve as the background for DM halo evolution. This impact can be examined through the shape and spin orientation of halos in numerical simulations incorporating IDE effects. In our work, we use the N-body simulation pipeline ME-GADGET to simulate and study the halo spin and orientation in IDE models. We found that in models where DM transfers into DE (IDE I), the alignment of halo shapes with the surrounding tidal field is enhanced, while the alignment of halo spins with the tidal field is decreased compared to $Λ$CDM. Conversely, in models where DE transfers into DM (IDE II), the opposite occurs. We have provided fitted functions to describe these alignment signals. Our study provides the foundation for more accurate modeling of observations in the future such as China Space Station Telescope.

astro-ph.CO↗

arXiv2025

Stratospheric Grid: A Wireless Power Transfer Enabled HAP Network with Integrated Generation-Grid-Load-Storage Functions

Conventional high-altitude platforms (HAPs) face challenges in achieving continuous all-weather operation due to intermittent photovoltaic power generation, limited energy storage capacity, and high mission loads resulting from functional integration. To address this fundamental issue, we propose a stratospheric energy grid in which wireless power transfer (WPT) interconnections constitute the grid layer, while HAPs operate as dynamically reconfigurable integrated generation-grid-load-storage (IGGLS) nodes that harvest, buffer, consume, and peer-to-peer transfer energy for constellation-level balancing and resilience. In this system, each HAP node can flexibly switch among energy source, load, and storage roles according to its energy status and mission requirements, enabling energy exchange and spatiotemporal optimization within the stratosphere. Through cooperative scheduling, the stratospheric grid not only enables surplus-to-deficit energy support among HAPs but also extends upward to satellites and downward to the terrestrial grid and communication infrastructure, forming a heterogeneous, WPT-mediated interconnection. As an IGGLS ecosystem, it exploits peer-to-peer energy logistics, spatiotemporal smoothing of intermittency, cross-domain backup via the terrestrial grid, and service-aware dispatch, thereby boosting overall energy utilization and operational resilience. The proposed approach is validated through case studies, and we delineate an agenda of feasible research directions.

eess.SY↗

arXiv2025

Imagine in Space: Exploring the Frontier of Spatial Intelligence and Reasoning Efficiency in Vision Language Models

Large language models (LLMs) and vision language models (VLMs), such as DeepSeek R1,OpenAI o3, and Gemini 2.5 Pro, have demonstrated remarkable reasoning capabilities across logical inference, problem solving, and decision making. However, spatial reasoning:a fundamental component of human cognition that includes mental rotation, navigation, and spatial relationship comprehension remains a significant challenge for current advanced VLMs. We hypothesize that imagination, the internal simulation of spatial states, is the dominant reasoning mechanism within a spatial world model. To test this hypothesis and systematically probe current VLM spatial reasoning mechanisms, we introduce SpatiaLite, a fully synthetic benchmark that jointly measures spatial reasoning accuracy and reasoning efficiency. Comprehensive experiments reveal three key findings. First, advanced VLMs predominantly rely on linguistic representations for reasoning and imagination, resulting in significant deficiencies on visual centric tasks that demand perceptual spatial relations and 3D geometry transformations such as mental rotation or projection prediction. Second, advanced VLMs exhibit severe inefficiency in their current spatial reasoning mechanisms, with token usage growing rapidly as transformation complexity increases. Third, we propose an Imagery Driven Framework (IDF) for data synthesis and training, which can implicitly construct an internal world model that is critical for spatial reasoning in VLMs. Building on SpatiaLite, this work delineates the spatial reasoning limits and patterns of advanced VLMs, identifies key shortcomings, and informs future advances

cs.AI↗

arXiv2025

A Global Geometric Analysis of Maximal Coding Rate Reduction

The maximal coding rate reduction (MCR$^2$) objective for learning structured and compact deep representations is drawing increasing attention, especially after its recent usage in the derivation of fully explainable and highly effective deep network architectures. However, it lacks a complete theoretical justification: only the properties of its global optima are known, and its global landscape has not been studied. In this work, we give a complete characterization of the properties of all its local and global optima, as well as other types of critical points. Specifically, we show that each (local or global) maximizer of the MCR$^2$ problem corresponds to a low-dimensional, discriminative, and diverse representation, and furthermore, each critical point of the objective is either a local maximizer or a strict saddle point. Such a favorable landscape makes MCR$^2$ a natural choice of objective for learning diverse and discriminative representations via first-order optimization methods. To validate our theoretical findings, we conduct extensive experiments on both synthetic and real data sets.

cs.LG↗

arXiv2025

From Structure to Detail: Hierarchical Distillation for Efficient Diffusion Model

The inference latency of diffusion models remains a critical barrier to their real-time application. While trajectory-based and distribution-based step distillation methods offer solutions, they present a fundamental trade-off. Trajectory-based methods preserve global structure but act as a "lossy compressor", sacrificing high-frequency details. Conversely, distribution-based methods can achieve higher fidelity but often suffer from mode collapse and unstable training. This paper recasts them from independent paradigms into synergistic components within our novel Hierarchical Distillation (HD) framework. We leverage trajectory distillation not as a final generator, but to establish a structural ``sketch", providing a near-optimal initialization for the subsequent distribution-based refinement stage. This strategy yields an ideal initial distribution that enhances the ceiling of overall performance. To further improve quality, we introduce and refine the adversarial training process. We find standard discriminator structures are ineffective at refining an already high-quality generator. To overcome this, we introduce the Adaptive Weighted Discriminator (AWD), tailored for the HD pipeline. By dynamically allocating token weights, AWD focuses on local imperfections, enabling efficient detail refinement. Our approach demonstrates state-of-the-art performance across diverse tasks. On ImageNet $256\times256$, our single-step model achieves an FID of 2.26, rivaling its 250-step teacher. It also achieves promising results on the high-resolution text-to-image MJHQ benchmark, proving its generalizability. Our method establishes a robust new paradigm for high-fidelity, single-step diffusion models.

cs.CV↗