arXiv subjects

Qian Zhang

Publications and source records attributed to Qian Zhang.

At least 199 records · Page 11Linked to original sources

arXiv2025

OmniMamba: Efficient and Unified Multimodal Understanding and Generation via State Space Models

Recent advancements in unified multimodal understanding and visual generation (or multimodal generation) models have been hindered by their quadratic computational complexity and dependence on large-scale training data. We present OmniMamba, the first linear-architecture-based multimodal generation model that generates both text and images through a unified next-token prediction paradigm. The model fully leverages Mamba-2's high computational and memory efficiency, extending its capabilities from text generation to multimodal generation. To address the data inefficiency of existing unified models, we propose two key innovations: (1) decoupled vocabularies to guide modality-specific generation, and (2) task-specific LoRA for parameter-efficient adaptation. Furthermore, we introduce a decoupled two-stage training strategy to mitigate data imbalance between two tasks. Equipped with these techniques, OmniMamba achieves competitive performance with JanusFlow while surpassing Show-o across benchmarks, despite being trained on merely 2M image-text pairs, which is 1,000 times fewer than Show-o. Notably, OmniMamba stands out with outstanding inference efficiency, achieving up to a 119.2 times speedup and 63% GPU memory reduction for long-sequence generation compared to Transformer-based counterparts. Code and models are released at https://github.com/hustvl/OmniMamba

cs.CV↗

arXiv2025

AlphaDrive: Unleashing the Power of VLMs in Autonomous Driving via Reinforcement Learning and Reasoning

OpenAI o1 and DeepSeek R1 achieve or even surpass human expert-level performance in complex domains like mathematics and science, with reinforcement learning (RL) and reasoning playing a crucial role. In autonomous driving, recent end-to-end models have greatly improved planning performance but still struggle with long-tailed problems due to limited common sense and reasoning abilities. Some studies integrate vision-language models (VLMs) into autonomous driving, but they typically rely on pre-trained models with simple supervised fine-tuning (SFT) on driving data, without further exploration of training strategies or optimizations specifically tailored for planning. In this paper, we propose AlphaDrive, a RL and reasoning framework for VLMs in autonomous driving. AlphaDrive introduces four GRPO-based RL rewards tailored for planning and employs a two-stage planning reasoning training strategy that combines SFT with RL. As a result, AlphaDrive significantly improves both planning performance and training efficiency compared to using only SFT or without reasoning. Moreover, we are also excited to discover that, following RL training, AlphaDrive exhibits some emergent multimodal planning capabilities, which is critical for improving driving safety and efficiency. To the best of our knowledge, AlphaDrive is the first to integrate GRPO-based RL with planning reasoning into autonomous driving. Code will be released to facilitate future research.

cs.CV↗

arXiv2025

Comparative high-pressure structural and electrical transport properties study of thermoelectric (Bi1-xSbx)2Te3 compounds

Thermoelectric (Bi1-x Sbx)2Te3 (BST-x) compounds with x=0.2, 0.7 and 0.9 have been studied using synchrotron angle-dispersive powder x-ray diffraction in a diamond anvil cell up to 25 GPa (at room temperature). The results clearly indicate that all compounds of this study follow a similar structural evolution with the one of pure Bi2Te3 and Sb2Te3 under pressure. From the comparison between the critical pressures of the corresponding phase transitions, a clear trend of increasing critical pressure for the transition to the disordered solid-solution BCC phase was observed with the increase of Sb concentration. In the case of the BST-0.7, an extended stability of the solid-solution BCC phase up to, at least, 180 GPa was observed. Finally, electrical transport properties measurements under pressure for BST-0.7, document a reversible pressure-induced metallization above 12 GPa.

cond-mat.mtrl-sci↗

arXiv2025

PediaBench: A Comprehensive Chinese Pediatric Dataset for Benchmarking Large Language Models

The emergence of Large Language Models (LLMs) in the medical domain has stressed a compelling need for standard datasets to evaluate their question-answering (QA) performance. Although there have been several benchmark datasets for medical QA, they either cover common knowledge across different departments or are specific to another department rather than pediatrics. Moreover, some of them are limited to objective questions and do not measure the generation capacity of LLMs. Therefore, they cannot comprehensively assess the QA ability of LLMs in pediatrics. To fill this gap, we construct PediaBench, the first Chinese pediatric dataset for LLM evaluation. Specifically, it contains 4,117 objective questions and 1,632 subjective questions spanning 12 pediatric disease groups. It adopts an integrated scoring criterion based on different difficulty levels to thoroughly assess the proficiency of an LLM in instruction following, knowledge understanding, clinical case analysis, etc. Finally, we validate the effectiveness of PediaBench with extensive experiments on 20 open-source and commercial LLMs. Through an in-depth analysis of experimental results, we offer insights into the ability of LLMs to answer pediatric questions in the Chinese context, highlighting their limitations for further improvements. Our code and data are published at https://github.com/ACMISLab/PediaBench.

cs.CL↗

arXiv2025

LLM-EvRep: Learning an LLM-Compatible Event Representation Using a Self-Supervised Framework

Recent advancements in event-based recognition have demonstrated significant promise, yet most existing approaches rely on extensive training, limiting their adaptability for efficient processing of event-driven visual content. Meanwhile, large language models (LLMs) have exhibited remarkable zero-shot capabilities across diverse domains, but their application to event-based visual recognition remains largely unexplored. To bridge this gap, we propose \textbf{LLM-EvGen}, an event representation generator that produces LLM-compatible event representations \textbf{LLM-EvRep}, thereby enhancing the performance of LLMs on event recognition tasks. The generator is trained using a self-supervised framework, aligning the generated representations with semantic consistency and structural fidelity. Comprehensive experiments were conducted on three datasets: N-ImageNet, N-Caltech101, and N-MNIST. The results demonstrate that our method, \textbf{LLM-EvRep}, outperforms the event-to-video method, E2VID, by 15.93\%, 0.82\%, and 50.21\%, respectively, in recognition tasks when evaluated using GPT-4o.

cs.CV↗

arXiv2025

OccRWKV: Rethinking Efficient 3D Semantic Occupancy Prediction with Linear Complexity

3D semantic occupancy prediction networks have demonstrated remarkable capabilities in reconstructing the geometric and semantic structure of 3D scenes, providing crucial information for robot navigation and autonomous driving systems. However, due to their large overhead from dense network structure designs, existing networks face challenges balancing accuracy and latency. In this paper, we introduce OccRWKV, an efficient semantic occupancy network inspired by Receptance Weighted Key Value (RWKV). OccRWKV separates semantics, occupancy prediction, and feature fusion into distinct branches, each incorporating Sem-RWKV and Geo-RWKV blocks. These blocks are designed to capture long-range dependencies, enabling the network to learn domain-specific representation (i.e., semantics and geometry), which enhances prediction accuracy. Leveraging the sparse nature of real-world 3D occupancy, we reduce computational overhead by projecting features into the bird's-eye view (BEV) space and propose a BEV-RWKV block for efficient feature enhancement and fusion. This enables real-time inference at 22.2 FPS without compromising performance. Experiments demonstrate that OccRWKV outperforms the state-of-the-art methods on the SemanticKITTI dataset, achieving a mIoU of 25.1 while being 20 times faster than the best baseline, Co-Occ, making it suitable for real-time deployment on robots to enhance autonomous navigation efficiency. Code and video are available on our project page: https://jmwang0117.github.io/OccRWKV/.

cs.CV↗

arXiv2025

Pushing up to the Limit of Memory Bandwidth and Capacity Utilization for Efficient LLM Decoding on Embedded FPGA

The extremely high computational and storage demands of large language models have excluded most edge devices, which were widely used for efficient machine learning, from being viable options. A typical edge device usually only has 4GB of memory capacity and a bandwidth of less than 20GB/s, while a large language model quantized to 4-bit precision with 7B parameters already requires 3.5GB of capacity, and its decoding process is purely bandwidth-bound. In this paper, we aim to explore these limits by proposing a hardware accelerator for large language model (LLM) inference on the Zynq-based KV260 platform, equipped with 4GB of 64-bit 2400Mbps DDR4 memory. We successfully deploy a LLaMA2-7B model, achieving a decoding speed of around 5 token/s, utilizing 93.3% of the memory capacity and reaching 85% decoding speed of the theoretical memory bandwidth limit. To fully reserve the memory capacity for model weights and key-value cache, we develop the system in a bare-metal environment without an operating system. To fully reserve the bandwidth for model weight transfers, we implement a customized dataflow with an operator fusion pipeline and propose a data arrangement format that can maximize the data transaction efficiency. This research marks the first attempt to deploy a 7B level LLM on a standalone embedded field programmable gate array (FPGA) device. It provides key insights into efficient LLM inference on embedded FPGA devices and provides guidelines for future architecture design.

cs.AR↗

arXiv2025

High pressure structural and lattice dynamics study of α-In$_2$Se$_3$

Layered $α$-In$_2$Se$_3$has been studied using a concomitant in-situ synchrotron angle dispersive powder x-ray diffraction and Raman spectroscopy study in a diamond anvil cell up to 60+ GPa, at room temperature. Helium, that remains fairly hydrostatic up to the highest pressure in this study, was used as the pressure-transmitting medium. The results from both experimental methods reveal a pressure-induced structural phase transition from $α$-In$_2$Se$_3$ to a monoclinic $β$'-In2Se3 structure at $\approx$1 GPa, in agreement with previous studies. Based on our detailed measurements using both experimental techniques and F-f formalism, the $β$'-In$_2$Se$_3$ structure remains stable up to 45 GPa, without a clear indication of a phase transition towards the previously reported $β$-In2Se3 phase. Above this pressure, In$_2$Se$_3$ adopts a disordered solid-solution-like orthorhombic structure, phase IV. The results are discussed in comparison with the relevant previous studies of $α$-In$_2$Se$_3$ under pressure.

cond-mat.mtrl-sci↗

arXiv2025

Molecular optomechanically-induced transparency

Molecular cavity optomechanics (COM), characterized by remarkably efficient optomechanical coupling enabled by a highly localized light field and ultra-small effective mode volume, holds significant promise for advancing applications in quantum science and technology. Here, we study optomechanically induced transparency and the associated group delay in a hybrid molecular COM system. We find that even with an extremely low optical quality factor, an obvious transparency window can appear, which is otherwise unattainable in a conventional COM system. Furthermore, by varying the ports of the probe light, the optomechanically induced transparency or absorption can be achieved, along with corresponding slowing or advancing of optical signals. These results indicate that our scheme provides a new method for adjusting the storage and retrieval of optical signals in such a molecular COM device.

physics.optics↗

arXiv2025

Generalizable Prompt Tuning for Vision-Language Models

Prompt tuning for vision-language models such as CLIP involves optimizing the text prompts used to generate image-text pairs for specific downstream tasks. While hand-crafted or template-based prompts are generally applicable to a wider range of unseen classes, they tend to perform poorly in downstream tasks (i.e., seen classes). Learnable soft prompts, on the other hand, often perform well in downstream tasks but lack generalizability. Additionally, prior research has predominantly concentrated on the textual modality, with very few studies attempting to explore the prompt's generalization potential from the visual modality. Keeping these limitations in mind, we investigate how to prompt tuning to obtain both a competitive downstream performance and generalization. The study shows that by treating soft and hand-crafted prompts as dual views of the textual modality, and maximizing their mutual information, we can better ensemble task-specific and general semantic information. Moreover, to generate more expressive prompts, the study introduces a class-wise augmentation from the visual modality, resulting in significant robustness to a wider range of unseen classes. Extensive evaluations on several benchmarks report that the proposed approach achieves competitive results in terms of both task-specific performance and general abilities.

cs.CV↗

arXiv2025

Operator Learning for Reconstructing Flow Fields from Sparse Measurements: an Energy Transformer Approach

Machine learning methods have shown great success in various scientific areas, including fluid mechanics. However, reconstruction problems, where full velocity fields must be recovered from partial observations, remain challenging. In this paper, we propose a novel operator learning framework for solving reconstruction problems by using the Energy Transformer (ET), an architecture inspired by associative memory models. We formulate reconstruction as a mapping from incomplete observed data to full reconstructed fields. The method is validated on three fluid mechanics examples using diverse types of data: (1) unsteady 2D vortex street in flow past a cylinder using simulation data; (2) high-speed under-expanded impinging supersonic jets impingement using Schlieren imaging; and (3) 3D turbulent jet flow using particle tracking. The results demonstrate the ability of ET to accurately reconstruct complex flow fields from highly incomplete data (90\% missing), even for noisy experimental measurements, with fast training and inference on a single GPU. This work provides a promising new direction for tackling reconstruction problems in fluid mechanics and other areas in mechanics, geophysics, weather prediction, and beyond.

physics.flu-dyn↗

arXiv2024

3D Registration in 30 Years: A Survey

3D point cloud registration is a fundamental problem in computer vision, computer graphics, robotics, remote sensing, and etc. Over the last thirty years, we have witnessed the amazing advancement in this area with numerous kinds of solutions. Although a handful of relevant surveys have been conducted, their coverage is still limited. In this work, we present a comprehensive survey on 3D point cloud registration, covering a set of sub-areas such as pairwise coarse registration, pairwise fine registration, multi-view registration, cross-scale registration, and multi-instance registration. The datasets, evaluation metrics, method taxonomy, discussions of the merits and demerits, insightful thoughts of future directions are comprehensively presented in this survey. The regularly updated project page of the survey is available at https://github.com/Amyyyy11/3D-Registration-in-30-Years-A-Survey.

cs.CV↗

arXiv2024

ShiftedBronzes: Benchmarking and Analysis of Domain Fine-Grained Classification in Open-World Settings

In real-world applications across specialized domains, addressing complex out-of-distribution (OOD) challenges is a common and significant concern. In this study, we concentrate on the task of fine-grained bronze ware dating, a critical aspect in the study of ancient Chinese history, and developed a benchmark dataset named ShiftedBronzes. By extensively expanding the bronze Ding dataset, ShiftedBronzes incorporates two types of bronze ware data and seven types of OOD data, which exhibit distribution shifts commonly encountered in bronze ware dating scenarios. We conduct benchmarking experiments on ShiftedBronzes and five commonly used general OOD datasets, employing a variety of widely adopted post-hoc, pre-trained Vision Large Model (VLM)-based and generation-based OOD detection methods. Through analysis of the experimental results, we validate previous conclusions regarding post-hoc, VLM-based, and generation-based methods, while also highlighting their distinct behaviors on specialized datasets. These findings underscore the unique challenges of applying general OOD detection methods to domain-specific tasks such as bronze ware dating. We hope that the ShiftedBronzes benchmark provides valuable insights into both the field of bronze ware dating and the and the development of OOD detection methods. The dataset and associated code will be available later.

cs.CV↗

arXiv2024

OTA-Key: Over the Air Key Management for Flexible and Reliable IoT Device Provision

As the Internet of Things (IoT) industry advances, the imperative to secure IoT devices has become increasingly critical. Current practices in both industry and academia advocate for the enhancement of device security through key installation. However, it has been observed that, in practice, IoT vendors frequently assign shared keys to batches of devices. This practice can expose devices to risks, such as data theft by attackers or large-scale Distributed Denial of Service (DDoS) attacks. To address this issue, our intuition is to assign a unique key to each device. Unfortunately, this strategy proves to be highly complex within the IoT context, as existing keys are typically hardcoded into the firmware, necessitating the creation of bespoke firmware for each device. Furthermore, correct pairing of device keys with their respective devices is crucial. Errors in this pairing process would incur substantial human and temporal resources to rectify and require extensive communication between IoT vendors, device manufacturers, and cloud platforms, leading to significant communication overhead. To overcome these challenges, we propose the OTA-Key scheme. This approach fundamentally decouples device keys from the firmware features stored in flash memory, utilizing an intermediary server to allocate unique device keys in two distinct stages and update keys. We conducted a formal security verification of our scheme using ProVerif and assessed its performance through a series of evaluations. The results demonstrate that our scheme is secure and effectively manages the large-scale distribution and updating of unique device keys. Additionally, it achieves significantly lower update times and data transfer volumes compared to other schemes.

cs.CR↗

arXiv2024

An Efficient Sum-Rate Maximization Algorithm for Fluid Antenna-Assisted ISAC System

This letter investigates a fluid antenna (FA)-assisted integrated sensing and communication (ISAC) system, with joint antenna position optimization and waveform design. We consider enhancing the sum-rate maximization (SRM) and sensing performance with the aid of FAs. Although the introduction of FAs brings more degrees of freedom for performance optimization, its position optimization poses a non-convex programming problem and brings great computational challenges. This letter contributes to building an efficient design algorithm by the block successive upper bound minimization and majorization-minimization principles, with each step admitting closed-form update for the ISAC waveform design. In addition, the extrapolation technique is exploited further to speed up the empirical convergence of FA position design. Simulation results show that the proposed design can achieve state-of-the-art sum-rate performance with at least 60% computation cutoff compared to existing works with successive convex approximation (SCA) and particle swarm optimization (PSO) algorithms.

cs.IT↗

arXiv2024

PerfGen: Automated Performance Benchmark Generation for Big Data Analytics

Many symptoms of poor performance in big data analytics such as computational skews, data skews, and memory skews are input dependent. However, due to the lack of inputs that can trigger such performance symptoms, it is hard to debug and test big data analytics. We design PerfGen to automatically generate inputs for the purpose of performance testing. PerfGen overcomes three challenges when naively using automated fuzz testing for the purpose of performance testing. First, typical greybox fuzzing relies on coverage as a guidance signal and thus is unlikely to trigger interesting performance behavior. Therefore, PerfGen provides performance monitor templates that a user can extend to serve as a set of guidance metrics for grey-box fuzzing. Second, performance symptoms may occur at an intermediate or later stage of a big data analytics pipeline. Thus, PerfGen uses a phased fuzzing approach. This approach identifies symptom-causing intermediate inputs at an intermediate stage first and then converts them to the inputs at the beginning of the program with a pseudo-inverse function generated by a large language model. Third, PerfGen defines sets of skew-inspired input mutations, which increases the chance of inducing performance problems. We evaluate PerfGen using four case studies. PerfGen achieves at least 11x speedup compared to a traditional fuzzing approach when generating inputs to trigger performance symptoms. Additionally, identifying intermediate inputs first and then converting them to original inputs enables PerfGen to generate such workloads in less than 0.004% of the iterations required by a baseline approach.

cs.SE↗

arXiv2024

Electronic-vibrational dynamics and coherence in x-ray transient absorption of N2+ induced by strong-field ionization

Attosecond transient absorption spectroscopy (ATAS) is becoming an indispensable and powerful tool in the emerging field of attochemistry, while the interpretation of measurements often requires full considerations of the coupling among various freedoms of motion. Here we develop the ionization-coupling model to incorporate the transient absorption and explore the coupled electronic-vibrational dynamics of nitrogen ions induced by strong-field ionization (SFI), which has been investigated in the recent transient x-ray K-edge absorption experiment [PRL 129, 123002 (2022)]. It is found the coherent vibrational wave packet on the involved electronic state is created with broad distribution of vibrational levels which leads to the spectral overlap on the K-edge absorption. By identifying the contributions of each electronic state, the study provides a different interpretation revealing the significant role of the excited state $A^2Π_u$ arising from the electronic-vibrational coupling in strong laser fields. We uncover new features of absorption from forbidden transitions during the laser pulse and confirm the vibronic coherence induced modulations of absorbance after SFI. A new scheme is proposed to avoid the spectral overlap and determine the population among the states which is crucial to resolve the debate on nitrogen air lasing. The work lays down the framework to research the ionic coherence in ATAS and offers valuable insights into the intricate interplay between electronic and vibrational dynamics.

physics.optics↗

arXiv2024

Vision Mamba: Efficient Visual Representation Learning with Bidirectional State Space Model

Recently the state space models (SSMs) with efficient hardware-aware designs, i.e., the Mamba deep learning model, have shown great potential for long sequence modeling. Meanwhile building efficient and generic vision backbones purely upon SSMs is an appealing direction. However, representing visual data is challenging for SSMs due to the position-sensitivity of visual data and the requirement of global context for visual understanding. In this paper, we show that the reliance on self-attention for visual representation learning is not necessary and propose a new generic vision backbone with bidirectional Mamba blocks (Vim), which marks the image sequences with position embeddings and compresses the visual representation with bidirectional state space models. On ImageNet classification, COCO object detection, and ADE20k semantic segmentation tasks, Vim achieves higher performance compared to well-established vision transformers like DeiT, while also demonstrating significantly improved computation & memory efficiency. For example, Vim is 2.8$\times$ faster than DeiT and saves 86.8% GPU memory when performing batch inference to extract features on images with a resolution of 1248$\times$1248. The results demonstrate that Vim is capable of overcoming the computation & memory constraints on performing Transformer-style understanding for high-resolution images and it has great potential to be the next-generation backbone for vision foundation models. Code is available at https://github.com/hustvl/Vim.

cs.CV↗