arXiv subjects

Peng Wang

Publications and source records attributed to Peng Wang.

At least 361 records · Page 20Linked to original sources

arXiv2025

MMCR: Advancing Visual Language Model in Multimodal Multi-Turn Contextual Reasoning

Compared to single-turn dialogue, multi-turn dialogue involving multiple images better aligns with the needs of real-world human-AI interactions. Additionally, as training data, it provides richer contextual reasoning information, thereby guiding the model to achieve better performance. However, existing vision-language models (VLMs) primarily rely on single-turn dialogue training and evaluation benchmarks. In this paper, following the characteristics of human dialogue, such as focused topics and concise, clear content, we present MMCR (Multimodal Multi-turn Contextual Reasoning), a novel dataset comprising: (1) MMCR-310k -- the largest multi-image multi-turn instruction tuning dataset with 310K contextual dialogues, each covering 1-4 images and 4 or 8 dialogue turns; and (2) MMCR-Bench -- a diagnostic benchmark featuring dialogues, spanning 8 domains (Humanities, Natural, Science, Education, etc.) and 40 sub-topics. Extensive evaluations demonstrate that models fine-tuned with MMCR-310k achieve 5.2\% higher contextual accuracy on MMCR-Bench, while showing consistent improvements on existing benchmarks (+1.1\% on AI2D, +1.2\% on MMMU and MMVet). MMCR and prompt engineering will be released publicly.

cs.AI↗

arXiv2025

DiffV2IR: Visible-to-Infrared Diffusion Model via Vision-Language Understanding

The task of translating visible-to-infrared images (V2IR) is inherently challenging due to three main obstacles: 1) achieving semantic-aware translation, 2) managing the diverse wavelength spectrum in infrared imagery, and 3) the scarcity of comprehensive infrared datasets. Current leading methods tend to treat V2IR as a conventional image-to-image synthesis challenge, often overlooking these specific issues. To address this, we introduce DiffV2IR, a novel framework for image translation comprising two key elements: a Progressive Learning Module (PLM) and a Vision-Language Understanding Module (VLUM). PLM features an adaptive diffusion model architecture that leverages multi-stage knowledge learning to infrared transition from full-range to target wavelength. To improve V2IR translation, VLUM incorporates unified Vision-Language Understanding. We also collected a large infrared dataset, IR-500K, which includes 500,000 infrared images compiled by various scenes and objects under various environmental conditions. Through the combination of PLM, VLUM, and the extensive IR-500K dataset, DiffV2IR markedly improves the performance of V2IR. Experiments validate DiffV2IR's excellence in producing high-quality translations, establishing its efficacy and broad applicability. The code, dataset, and DiffV2IR model will be available at https://github.com/LidongWang-26/DiffV2IR.

cs.CV↗

arXiv2025

A Shared Low-Rank Adaptation Approach to Personalized RLHF

Reinforcement Learning from Human Feedback (RLHF) has emerged as a pivotal technique for aligning artificial intelligence systems with human values, achieving remarkable success in fine-tuning large language models. However, existing RLHF frameworks often assume that human preferences are relatively homogeneous and can be captured by a single, unified reward model. This assumption overlooks the inherent diversity and heterogeneity across individuals, limiting the adaptability of RLHF to personalized scenarios and risking misalignments that can diminish user satisfaction and trust in AI systems. In this paper, we address these challenges by introducing Low-Rank Adaptation (LoRA) into the personalized RLHF framework. We apply LoRA in the the aggregated parameter space of all personalized reward functions, thereby enabling efficient learning of personalized reward models from potentially limited local datasets. Our approach exploits potential shared structures among the local ground-truth reward models while allowing for individual adaptation, without relying on restrictive assumptions about shared representations as in prior works. We further establish sample complexity guarantees for our method. Theoretical analysis demonstrates the effectiveness of the proposed approach in capturing both shared and individual-specific structures within heterogeneous human preferences, addressing the dual challenge of personalization requirements and practical data constraints. Experimental results on real-world datasets corroborate the efficiency of our algorithm in the personalized RLHF setting.

cs.LG↗

arXiv2025

New constructions of asymptotically optimal periodic and aperiodic quasi-complementary sequence sets

Quasi-complementary sequence sets (QCSSs) play an important role in multi-carrier code division multiple access (MC-CDMA) systems as they can support more users than perfect complementary sequence sets (PCSSs). The objective of this paper is to present new constructions of asymptotically optimal periodic and aperiodic QCSSs with large set sizes. Firstly, we construct a family of asymptotically optimal periodic $(p^{2n}, p^n-1, p^n-1, p^n+1)$ QCSSs with small alphabet size $p$, which has larger set size than the known family of periodic $(p^n(p^n-1), p^n-1, p^n-1, p^n+1)$ QCSSs. Secondly, we construct five new families of asymptotically optimal aperiodic QCSSs with large set sizes and low aperiodic tolerances. Each family of these aperiodic QCSSs has set size $Θ(K^2)$ for some flock size $K$. Compared with known asymptotically optimal aperiodic QCSSs in the literature, the proposed aperiodic QCSSs by us have better parameters or new lengths of their constituent sequences.

cs.IT↗

arXiv2025

Orthogonal Alignment of Galaxy Group Angular Momentum with Cosmic Filament Spines: An Observational Study

We investigate the alignment between the angular momenta of galaxy groups and the spines of their associated cosmic filaments. Our results demonstrate a significant tendency for these two orientations to be perpendicular, indicating that the rotation of a galaxy group does not originate from the spin of cosmic filaments. Instead, it is driven by the orbital angular momentum contributed by member galaxies as they accrete along the direction of the filament spines. Moreover, the strength of this perpendicular alignment signal varies with the richness of the galaxy groups, with the most pronounced alignment observed among the wealthiest groups. This pronounced alignment is largely due to the more coherent spatial distribution of member galaxies in richer groups relative to the filament spines. Our study provides valuable insights into the mechanisms of angular momentum acquisition in galaxy groups from an observational standpoint.

astro-ph.GA↗

arXiv2025

Unlocking Generalization Power in LiDAR Point Cloud Registration

In real-world environments, a LiDAR point cloud registration method with robust generalization capabilities (across varying distances and datasets) is crucial for ensuring safety in autonomous driving and other LiDAR-based applications. However, current methods fall short in achieving this level of generalization. To address these limitations, we propose UGP, a pruned framework designed to enhance generalization power for LiDAR point cloud registration. The core insight in UGP is the elimination of cross-attention mechanisms to improve generalization, allowing the network to concentrate on intra-frame feature extraction. Additionally, we introduce a progressive self-attention module to reduce ambiguity in large-scale scenes and integrate Bird's Eye View (BEV) features to incorporate semantic information about scene elements. Together, these enhancements significantly boost the network's generalization performance. We validated our approach through various generalization experiments in multiple outdoor scenes. In cross-distance generalization experiments on KITTI and nuScenes, UGP achieved state-of-the-art mean Registration Recall rates of 94.5% and 91.4%, respectively. In cross-dataset generalization from nuScenes to KITTI, UGP achieved a state-of-the-art mean Registration Recall of 90.9%. Code will be available at https://github.com/peakpang/UGP.

cs.CV↗

arXiv2025

Imaging neutron radiation-induced defects in single-crystal chemical vapor deposition diamond at the atomic level

Diamond's exceptional properties make it highly suited for applications in challenging radiation environments. Understanding radiation-induced damage in diamond is crucial for enabling its practical applications and advancing materials science. However, direct imaging of radiation-induced crystal defects at the atomic scale remains rare due to diamond's compact lattice structure. Here, we report the atomic-level characterization of crystal defects induced by high-flux fast neutron radiation (up to $3 \times10^{17}$ n/$cm^2$) in single-crystal chemical vapor deposition diamonds. Through Raman spectroscopy, the phase transition from carbon $sp^3$ to $sp^2$ hybridization was identified, primarily associated with the formation of dumbbell-shaped interstitial defects. Using electron energy loss spectroscopy and aberration-corrected transmission electron microscopy, we observed a clustering trend in defect distribution, where $sp^2$ rich clusters manifested as dislocation structures with a density up to $10^{14}$ $cm^{-2}$. Lomer-Cottrell junctions were identified, offering a possible explanation for defect cluster formation. Radiation-induced point defects were found to be dispersed throughout the diamond lattice, highlighting the widespread nature of primary defect formation. Vacancy defects, along with $\langle 111 \rangle$ and $\langle 100 \rangle$ oriented dumbbell-shaped interstitial defects induced by high-dose neutron irradiation, were directly imaged, providing microscopic structural evidence that complements spectroscopic studies of point defects. Dynamical simulations combined with an adiabatic recombination-based damage model provided insights into the correlation between irradiation dose and resulting crystal damage. These findings advance our understanding of neutron-induced damage mechanisms in diamond and contribute to the development of radiation-resistant diamond materials.

cond-mat.mtrl-sci↗

arXiv2025

Zero-Shot Subject-Centric Generation for Creative Application Using Entropy Fusion

Generative models are widely used in visual content creation. However, current text-to-image models often face challenges in practical applications-such as textile pattern design and meme generation-due to the presence of unwanted elements that are difficult to separate with existing methods. Meanwhile, subject-reference generation has emerged as a key research trend, highlighting the need for techniques that can produce clean, high-quality subject images while effectively removing extraneous components. To address this challenge, we introduce a framework for reliable subject-centric image generation. In this work, we propose an entropy-based feature-weighted fusion method to merge the informative cross-attention features obtained from each sampling step of the pretrained text-to-image model FLUX, enabling a precise mask prediction and subject-centric generation. Additionally, we have developed an agent framework based on Large Language Models (LLMs) that translates users' casual inputs into more descriptive prompts, leading to highly detailed image generation. Simultaneously, the agents extract primary elements of prompts to guide the entropy-based feature fusion, ensuring focused primary element generation without extraneous components. Experimental results and user studies demonstrate our methods generates high-quality subject-centric images, outperform existing methods or other possible pipelines, highlighting the effectiveness of our approach.

cs.CV↗

arXiv2025

Seedream 2.0: A Native Chinese-English Bilingual Image Generation Foundation Model

Rapid advancement of diffusion models has catalyzed remarkable progress in the field of image generation. However, prevalent models such as Flux, SD3.5 and Midjourney, still grapple with issues like model bias, limited text rendering capabilities, and insufficient understanding of Chinese cultural nuances. To address these limitations, we present Seedream 2.0, a native Chinese-English bilingual image generation foundation model that excels across diverse dimensions, which adeptly manages text prompt in both Chinese and English, supporting bilingual image generation and text rendering. We develop a powerful data system that facilitates knowledge integration, and a caption system that balances the accuracy and richness for image description. Particularly, Seedream is integrated with a self-developed bilingual large language model as a text encoder, allowing it to learn native knowledge directly from massive data. This enable it to generate high-fidelity images with accurate cultural nuances and aesthetic expressions described in either Chinese or English. Beside, Glyph-Aligned ByT5 is applied for flexible character-level text rendering, while a Scaled ROPE generalizes well to untrained resolutions. Multi-phase post-training optimizations, including SFT and RLHF iterations, further improve the overall capability. Through extensive experimentation, we demonstrate that Seedream 2.0 achieves state-of-the-art performance across multiple aspects, including prompt-following, aesthetics, text rendering, and structural correctness. Furthermore, Seedream 2.0 has been optimized through multiple RLHF iterations to closely align its output with human preferences, as revealed by its outstanding ELO score. In addition, it can be readily adapted to an instruction-based image editing model, such as SeedEdit, with strong editing capability that balances instruction-following and image consistency.

cs.CV↗

arXiv2025

The large-scale structure supplies the formation of gas-star misaligned galaxies

Using the integral field unit data from the Mapping Nearby Galaxies at Apache Point Observatory (MaNGA) survey, we build a sample of gas-star misaligned galaxies. The large-scale environment of misaligned galaxies is dominated by filaments and clusters, while is less dense relative to the gas-star aligned control galaxies. The direction of the large-scale structure (LSS) is defined by its minor axis ($\vec{e_{3}}$), which indicates the slowest collapsing direction. For the aligned controls, the gas and stellar spins are preferentially perpendicular to $\vec{e_{3}}$, since these galaxies reside in high-mass host haloes. For the misaligned galaxies, the gas spins also tend to be perpendicular to $\vec{e_{3}}$, suggesting that misaligned gas is recently accreted from the LSS. Meanwhile, there is no correlation between their stellar spins and $\vec{e_{3}}$. There are two possible explanations for this observational phenomenon: (1) the large-scale environments of misaligned galaxies evolve as they grow, with stellar angular momenta acquiring in different environments having different orientations; (2) the correlation between stellar spins and the LSS is smeared out since a relatively higher portion of misaligned galaxies in sheet environments are statistically analysed together with those in filament environments.

astro-ph.GA↗

arXiv2025

Fixing the AdS$_3$ metric from the pure state entanglement entropies of CFT$_2$

In this paper, based on RT formula, by identifying the pure state UV and IR entanglement entropies of a perturbed CFT$_2$ with geodesic lengths in the bulk, we demonstrate that the dual geometry is uniquely determined to be asymptotically AdS$_3$. The pure AdS$_3$ geometry is recovered by taking the massless limit of the system. Our derivations hold in both static and covariant scenarios.

hep-th↗

arXiv2025

Octopus: Alleviating Hallucination via Dynamic Contrastive Decoding

Large Vision-Language Models (LVLMs) have obtained impressive performance in visual content understanding and multi-modal reasoning. Unfortunately, these large models suffer from serious hallucination problems and tend to generate fabricated responses. Recently, several Contrastive Decoding (CD) strategies have been proposed to alleviate hallucination by introducing disturbed inputs. Although great progress has been made, these CD strategies mostly apply a one-size-fits-all approach for all input conditions. In this paper, we revisit this process through extensive experiments. Related results show that hallucination causes are hybrid and each generative step faces a unique hallucination challenge. Leveraging these meaningful insights, we introduce a simple yet effective Octopus-like framework that enables the model to adaptively identify hallucination types and create a dynamic CD workflow. Our Octopus framework not only outperforms existing methods across four benchmarks but also demonstrates excellent deployability and expansibility. Code is available at https://github.com/LijunZhang01/Octopus.

cs.CV↗

arXiv2025

ColorDynamic: Generalizable, Scalable, Real-time, End-to-end Local Planner for Unstructured and Dynamic Environments

Deep Reinforcement Learning (DRL) has demonstrated potential in addressing robotic local planning problems, yet its efficacy remains constrained in highly unstructured and dynamic environments. To address these challenges, this study proposes the ColorDynamic framework. First, an end-to-end DRL formulation is established, which maps raw sensor data directly to control commands, thereby ensuring compatibility with unstructured environments. Under this formulation, a novel network, Transqer, is introduced. The Transqer enables online DRL learning from temporal transitions, substantially enhancing decision-making in dynamic scenarios. To facilitate scalable training of Transqer with diverse data, an efficient simulation platform E-Sparrow, along with a data augmentation technique leveraging symmetric invariance, are developed. Comparative evaluations against state-of-the-art methods, alongside assessments of generalizability, scalability, and real-time performance, were conducted to validate the effectiveness of ColorDynamic. Results indicate that our approach achieves a success rate exceeding 90% while exhibiting real-time capacity (1.2-1.3 ms per planning). Additionally, ablation studies were performed to corroborate the contributions of individual components. Building on this, the OkayPlan-ColorDynamic (OPCD) navigation system is presented, with simulated and real-world experiments demonstrating its superiority and applicability in complex scenarios. The codebase and experimental demonstrations have been open-sourced on our website to facilitate reproducibility and further research.

cs.RO↗

arXiv2025

Brain-inspired analogical mixture prototypes for few-shot class-incremental learning

Few-shot class-incremental learning (FSCIL) poses significant challenges for artificial neural networks due to the need to efficiently learn from limited data while retaining knowledge of previously learned tasks. Inspired by the brain's mechanisms for categorization and analogical learning, we propose a novel approach called Brain-inspired Analogical Mixture Prototypes (BAMP). BAMP has three components: mixed prototypical feature learning, statistical analogy, and soft voting. Starting from a pre-trained Vision Transformer (ViT), mixed prototypical feature learning represents each class using a mixture of prototypes and fine-tunes these representations during the base session. The statistical analogy calibrates the mean and covariance matrix of prototypes for new classes according to similarity to the base classes, and computes classification score with Mahalanobis distance. Soft voting combines both merits of statistical analogy and an off-shelf FSCIL method. Our experiments on benchmark datasets demonstrate that BAMP outperforms state-of-the-art on both traditional big start FSCIL setting and challenging small start FSCIL setting. The study suggests that brain-inspired analogical mixture prototypes can alleviate catastrophic forgetting and over-fitting problems in FSCIL.

cs.CV↗

arXiv2025

CEB: Compositional Evaluation Benchmark for Fairness in Large Language Models

As Large Language Models (LLMs) are increasingly deployed to handle various natural language processing (NLP) tasks, concerns regarding the potential negative societal impacts of LLM-generated content have also arisen. To evaluate the biases exhibited by LLMs, researchers have recently proposed a variety of datasets. However, existing bias evaluation efforts often focus on only a particular type of bias and employ inconsistent evaluation metrics, leading to difficulties in comparison across different datasets and LLMs. To address these limitations, we collect a variety of datasets designed for the bias evaluation of LLMs, and further propose CEB, a Compositional Evaluation Benchmark that covers different types of bias across different social groups and tasks. The curation of CEB is based on our newly proposed compositional taxonomy, which characterizes each dataset from three dimensions: bias types, social groups, and tasks. By combining the three dimensions, we develop a comprehensive evaluation strategy for the bias in LLMs. Our experiments demonstrate that the levels of bias vary across these dimensions, thereby providing guidance for the development of specific bias mitigation methods.

cs.CL↗

arXiv2025

Dimensionality-Reduction Techniques for Approximate Nearest Neighbor Search: A Survey and Evaluation

Approximate Nearest Neighbor Search (ANNS) on high-dimensional vectors has become a fundamental and essential component in various machine learning tasks. Recently, with the rapid development of deep learning models and the applications of Large Language Models (LLMs), the dimensionality of the vectors keeps growing in order to accommodate a richer semantic representation. This poses a major challenge to the ANNS solutions since distance calculation cost in ANNS grows linearly with the dimensionality of vectors. To overcome this challenge, dimensionality-reduction techniques can be leveraged to accelerate the distance calculation in the search process. In this paper, we investigate six dimensionality-reduction techniques that have the potential to improve ANNS solutions, including classical algorithms such as PCA and vector quantization, as well as algorithms based on deep learning approaches. We further describe two frameworks to apply these techniques in the ANNS workflow, and theoretically analyze the time and space costs, as well as the beneficial threshold for the pruning ratio of these techniques. The surveyed techniques are evaluated on six public datasets. The analysis of the results reveals the characteristics of the different families of techniques and provides insights into the promising future research directions.

cs.DB↗

arXiv2025

Qwen2.5-VL Technical Report

We introduce Qwen2.5-VL, the latest flagship model of Qwen vision-language series, which demonstrates significant advancements in both foundational capabilities and innovative functionalities. Qwen2.5-VL achieves a major leap forward in understanding and interacting with the world through enhanced visual recognition, precise object localization, robust document parsing, and long-video comprehension. A standout feature of Qwen2.5-VL is its ability to localize objects using bounding boxes or points accurately. It provides robust structured data extraction from invoices, forms, and tables, as well as detailed analysis of charts, diagrams, and layouts. To handle complex inputs, Qwen2.5-VL introduces dynamic resolution processing and absolute time encoding, enabling it to process images of varying sizes and videos of extended durations (up to hours) with second-level event localization. This allows the model to natively perceive spatial scales and temporal dynamics without relying on traditional normalization techniques. By training a native dynamic-resolution Vision Transformer (ViT) from scratch and incorporating Window Attention, we reduce computational overhead while maintaining native resolution. As a result, Qwen2.5-VL excels not only in static image and document understanding but also as an interactive visual agent capable of reasoning, tool usage, and task execution in real-world scenarios such as operating computers and mobile devices. Qwen2.5-VL is available in three sizes, addressing diverse use cases from edge AI to high-performance computing. The flagship Qwen2.5-VL-72B model matches state-of-the-art models like GPT-4o and Claude 3.5 Sonnet, particularly excelling in document and diagram understanding. Additionally, Qwen2.5-VL maintains robust linguistic performance, preserving the core language competencies of the Qwen2.5 LLM.

cs.CV↗

arXiv2025

SeWA: Selective Weight Average via Probabilistic Masking

Weight averaging has become a standard technique for enhancing model performance. However, methods such as Stochastic Weight Averaging (SWA) and Latest Weight Averaging (LAWA) often require manually designed procedures to sample from the training trajectory, and the results depend heavily on hyperparameter tuning. To minimize human effort, this paper proposes a simple yet efficient algorithm called Selective Weight Averaging (SeWA), which adaptively selects checkpoints during the final stages of training for averaging. Based on SeWA, we show that only a few points are needed to achieve better generalization and faster convergence. Theoretically, solving the discrete subset selection problem is inherently challenging. To address this, we transform it into a continuous probabilistic optimization framework and employ the Gumbel-Softmax estimator to learn the non-differentiable mask for each checkpoint. Further, we theoretically derive the SeWA's stability-based generalization bounds, which are sharper than that of SGD under both convex and non-convex assumptions. Finally, solid extended experiments in various domains, including behavior cloning, image classification, and text classification, further validate the effectiveness of our approach.

cs.LG↗