arXiv subjects

Han Wang

Publications and source records attributed to Han Wang.

At least 163 records · Page 9Linked to original sources

arXiv2025

LSP-ST: Ladder Shape-Biased Side-Tuning for Robust Infrared Small Target Detection

Fine-tuning the Segment Anything Model (SAM) for infrared small target detection poses significant challenges due to severe domain shifts. Existing adaptation methods often incorporate handcrafted priors to bridge this gap, yet such designs limit generalization and scalability. We identify a fundamental texture bias in foundation models, which overly depend on local texture cues for target localization. To address this, we propose Ladder Shape-Biased Side-Tuning (LSP-ST), a novel approach that introduces a shape-aware inductive bias to facilitate effective adaptation beyond texture cues. In contrast to prior work that injects explicit edge or contour features, LSP-ST models shape as a global structural prior, integrating both boundaries and internal layouts. We design a Shape-Enhanced Large-Kernel Attention Module to hierarchically and implicitly capture structural information in a fully differentiable manner, without task-specific handcrafted guidance. A theoretical analysis grounded in matched filtering and backpropagation reveals the mechanism by which the proposed attention improves structure-aware learning. With only 4.72M learnable parameters, LSP-ST achieves state-of-the-art performance on multiple infrared small target detection benchmarks. Furthermore, its strong generalization is validated across tasks such as mirror detection, shadow detection, and camouflaged object detection, while maintaining stable performance on texture-driven tasks like salient object detection, demonstrating that the introduced shape bias complements rather than competes with texture-based reasoning.

cs.CV↗

arXiv2025

SUMMA: A Multimodal Large Language Model for Advertisement Summarization

Understanding multimodal video ads is crucial for improving query-ad matching and relevance ranking on short video platforms, enhancing advertising effectiveness and user experience. However, the effective utilization of multimodal information with high commercial value still largely constrained by reliance on highly compressed video embeddings-has long been inadequate. To address this, we propose SUMMA (the abbreviation of Summarizing MultiModal Ads), a multimodal model that automatically processes video ads into summaries highlighting the content of highest commercial value, thus improving their comprehension and ranking in Douyin search-advertising systems. SUMMA is developed via a two-stage training strategy-multimodal supervised fine-tuning followed by reinforcement learning with a mixed reward mechanism-on domain-specific data containing video frames and ASR/OCR transcripts, generating commercially valuable and explainable summaries. We integrate SUMMA-generated summaries into our production pipeline, directly enhancing the candidate retrieval and relevance ranking stages in real search-advertising systems. Both offline and online experiments show substantial improvements over baselines, with online results indicating a statistically significant 1.5% increase in advertising revenue. Our work establishes a novel paradigm for condensing multimodal information into representative texts, effectively aligning visual ad content with user query intent in retrieval and recommendation scenarios.

cs.IR↗

arXiv2025

DASICS White Paper: Enhancing Memory Protection with Dynamic Compartmentalization

In the existing software development ecosystem, security issues introduced by third-party code cannot be overlooked. Among these security concerns, memory access vulnerabilities stand out prominently, leading to risks such as the theft or tampering of sensitive data. To address this issue, software-based defense mechanisms have been established at the programming language, compiler, and operating system levels. However, as a trade-off, these mechanisms significantly reduce software execution efficiency. Hardware-software co-design approaches have sought to either construct entirely isolated trusted execution environments or attempt to partition security domains within the same address space. While such approaches enhance efficiency compared to pure software methods, they also encounter challenges related to granularity of protection, performance overhead, and portability. In response to these challenges, we present the DASICS (Dynamic in-Address-Space Isolation by Code Segments) secure processor design, which offers dynamic and flexible security protection across multiple privilege levels, addressing data flow protection, control flow protection, and secure system calls. We have implemented hardware FPGA prototypes and software QEMU simulator prototypes based on DASICS, along with necessary modifications to system software for adaptability. We illustrate the protective mechanisms and effectiveness of DASICS with two practical examples and provide potential real-world use cases where DASICS could be applied.

cs.CR↗

arXiv2025

FreeInsert: Personalized Object Insertion with Geometric and Style Control

Text-to-image diffusion models have made significant progress in image generation, allowing for effortless customized generation. However, existing image editing methods still face certain limitations when dealing with personalized image composition tasks. First, there is the issue of lack of geometric control over the inserted objects. Current methods are confined to 2D space and typically rely on textual instructions, making it challenging to maintain precise geometric control over the objects. Second, there is the challenge of style consistency. Existing methods often overlook the style consistency between the inserted object and the background, resulting in a lack of realism. In addition, the challenge of inserting objects into images without extensive training remains significant. To address these issues, we propose \textit{FreeInsert}, a novel training-free framework that customizes object insertion into arbitrary scenes by leveraging 3D geometric information. Benefiting from the advances in existing 3D generation models, we first convert the 2D object into 3D, perform interactive editing at the 3D level, and then re-render it into a 2D image from a specified view. This process introduces geometric controls such as shape or view. The rendered image, serving as geometric control, is combined with style and content control achieved through diffusion adapters, ultimately producing geometrically controlled, style-consistent edited images via the diffusion model.

cs.CV↗

arXiv2025

Early Prediction of In-Hospital ICU Mortality Using Innovative First-Day Data: A Review

The intensive care unit (ICU) manages critically ill patients, many of whom face a high risk of mortality. Early and accurate prediction of in-hospital mortality within the first 24 hours of ICU admission is crucial for timely clinical interventions, resource optimization, and improved patient outcomes. Traditional scoring systems, while useful, often have limitations in predictive accuracy and adaptability. Objective: This review aims to systematically evaluate and benchmark innovative methodologies that leverage data available within the first day of ICU admission for predicting in-hospital mortality. We focus on advancements in machine learning, novel biomarker applications, and the integration of diverse data types.

cs.LG↗

arXiv2025

SAIL-VL2 Technical Report

We introduce SAIL-VL2, an open-suite vision-language foundation model (LVM) for comprehensive multimodal understanding and reasoning. As the successor to SAIL-VL, SAIL-VL2 achieves state-of-the-art performance at the 2B and 8B parameter scales across diverse image and video benchmarks, demonstrating strong capabilities from fine-grained perception to complex reasoning. Its effectiveness is driven by three core innovations. First, a large-scale data curation pipeline with scoring and filtering strategies enhances both quality and distribution across captioning, OCR, QA, and video data, improving training efficiency. Second, a progressive training framework begins with a powerful pre-trained vision encoder (SAIL-ViT), advances through multimodal pre-training, and culminates in a thinking-fusion SFT-RL hybrid paradigm that systematically strengthens model capabilities. Third, architectural advances extend beyond dense LLMs to efficient sparse Mixture-of-Experts (MoE) designs. With these contributions, SAIL-VL2 demonstrates competitive performance across 106 datasets and achieves state-of-the-art results on challenging reasoning benchmarks such as MMMU and MathVista. Furthermore, on the OpenCompass leaderboard, SAIL-VL2-2B ranks first among officially released open-source models under the 4B parameter scale, while serving as an efficient and extensible foundation for the open-source multimodal community.

cs.CV↗

arXiv2025

Efficient lattice field theory simulation using adaptive normalizing flow on a resistive memory-based neural differential equation solver

Lattice field theory (LFT) simulations underpin advances in classical statistical mechanics and quantum field theory, providing a unified computational framework across particle, nuclear, and condensed matter physics. However, the application of these methods to high-dimensional systems remains severely constrained by several challenges, including the prohibitive computational cost and limited parallelizability of conventional sampling algorithms such as hybrid Monte Carlo (HMC), the substantial training expense associated with traditional normalizing flow models, and the inherent energy inefficiency of digital hardware architectures. Here, we introduce a software-hardware co-design that integrates an adaptive normalizing flow (ANF) model with a resistive memory-based neural differential equation solver, enabling efficient generation of LFT configurations. Software-wise, ANF enables efficient parallel generation of statistically independent configurations, thereby reducing computational costs, while low-rank adaptation (LoRA) allows cost-effective fine-tuning across diverse simulation parameters. Hardware-wise, in-memory computing with resistive memory substantially enhances both parallelism and energy efficiency. We validate our approach on the scalar phi4 theory and the effective field theory of graphene wires, using a hybrid analog-digital neural differential equation solver equipped with a 180 nm resistive memory in-memory computing macro. Our co-design enables low-cost computation, achieving approximately 8.2-fold and 13.9-fold reductions in integrated autocorrelation time over HMC, while requiring fine-tuning of less than 8% of the weights via LoRA. Compared to state-of-the-art GPUs, our co-design achieves up to approximately 16.1- and 17.0-fold speedups for the two tasks, as well as 73.7- and 138.0-fold improvements in energy efficiency.

cs.NE↗

arXiv2025

Combinatorial optimization enhanced by shallow quantum circuits with 104 superconducting qubits

A pivotal task for quantum computing is to speed up solving problems that are both classically intractable and practically valuable. Among these, combinatorial optimization problems have attracted tremendous attention due to their broad applicability and natural fitness to Ising Hamiltonians. Here we propose a quantum sampling strategy, based on which we design an algorithm for accelerating solving the ground states of Ising model, a class of NP-hard problems in combinatorial optimization. The algorithm employs a hybrid quantum-classical workflow, with a shallow-circuit quantum sampling subroutine dedicated to navigating the energy landscape. Using up to 104 superconducting qubits, we demonstrate that this algorithm outputs favorable solutions against even a highly-optimized classical simulated annealing (SA) algorithm. Furthermore, we illustrate the path toward quantum speedup based on the time-to-solution metric against SA running on a single-core CPU with just 100 qubits. Our results indicate a promising alternative to classical heuristics for combinatorial optimization, a paradigm where quantum advantage might become possible on near-term superconducting quantum processors with thousands of qubits and without the assistance of error correction.

quant-ph↗

arXiv2025

OpenCSP: A Deep Learning Framework for Crystal Structure Prediction from Ambient to High Pressure

High-pressure crystal structure prediction (CSP) underpins advances in condensed matter physics, planetary science, and materials discovery. Yet, most large atomistic models are trained on near-ambient, equilibrium data, leading to degraded stress accuracy at tens to hundreds of gigapascals and sparse coverage of pressure-stabilized stoichiometries and dense coordination motifs. Here, we introduce OpenCSP, a machine learning framework for CSP tasks spanning ambient to high-pressure conditions. This framework comprises an open-source pressure-resolved dataset alongside a suite of publicly available atomistic models that are jointly optimized for accuracy in energy, force, and stress predictions. The dataset is constructed via randomized high-pressure sampling and iteratively refined through an uncertainty-guided concurrent learning strategy, which enriches underrepresented compression regimes while suppressing redundant DFT labeling. Despite employing a training corpus one to two orders of magnitude smaller than those of leading large models, OpenCSP achieves comparable or superior performance in high-pressure enthalpy ranking and stability prediction. Across benchmark CSP tasks spanning a wide pressure window, our models match or surpass MACE-MPA-0, MatterSim v1 5M, and GRACE-2L-OAM, with the largest gains observed at elevated pressures. These results demonstrate that targeted, pressure-aware data acquisition coupled with scalable architectures enables data-efficient, high-fidelity CSP, paving the way for autonomous materials discovery under ambient and extreme conditions.

cond-mat.mtrl-sci↗

arXiv2025

Benchmarking Vision-Language Models on Chinese Ancient Documents: From OCR to Knowledge Reasoning

Chinese ancient documents, invaluable carriers of millennia of Chinese history and culture, hold rich knowledge across diverse fields but face challenges in digitization and understanding, i.e., traditional methods only scan images, while current Vision-Language Models (VLMs) struggle with their visual and linguistic complexity. Existing document benchmarks focus on English printed texts or simplified Chinese, leaving a gap for evaluating VLMs on ancient Chinese documents. To address this, we present AncientDoc, the first benchmark for Chinese ancient documents, designed to assess VLMs from OCR to knowledge reasoning. AncientDoc includes five tasks (page-level OCR, vernacular translation, reasoning-based QA, knowledge-based QA, linguistic variant QA) and covers 14 document types, over 100 books, and about 3,000 pages. Based on AncientDoc, we evaluate mainstream VLMs using multiple metrics, supplemented by a human-aligned large language model for scoring.

cs.CL↗

arXiv2025

PLRV-O: Advancing Differentially Private Deep Learning via Privacy Loss Random Variable Optimization

Differentially Private Stochastic Gradient Descent (DP-SGD) is a standard method for enforcing privacy in deep learning, typically using the Gaussian mechanism to perturb gradient updates. However, conventional mechanisms such as Gaussian and Laplacian noise are parameterized only by variance or scale. This single degree of freedom ties the magnitude of noise directly to both privacy loss and utility degradation, preventing independent control of these two factors. The problem becomes more pronounced when the number of composition rounds T and batch size B vary across tasks, as these variations induce task-dependent shifts in the privacy-utility trade-off, where small changes in noise parameters can disproportionately affect model accuracy. To address this limitation, we introduce PLRV-O, a framework that defines a broad search space of parameterized DP-SGD noise distributions, where privacy loss moments are tightly characterized yet can be optimized more independently with respect to utility loss. This formulation enables systematic adaptation of noise to task-specific requirements, including (i) model size, (ii) training duration, (iii) batch sampling strategies, and (iv) clipping thresholds under both training and fine-tuning settings. Empirical results demonstrate that PLRV-O substantially improves utility under strict privacy constraints. On CIFAR-10, a fine-tuned ViT achieves 94.03% accuracy at epsilon approximately 0.5, compared to 83.93% with Gaussian noise. On SST-2, RoBERTa-large reaches 92.20% accuracy at epsilon approximately 0.2, versus 50.25% with Gaussian.

cs.CR↗

arXiv2025

GWSpace: a multi-mission science data simulator for space-based gravitational wave detection

Space-based gravitational wave detectors such as TianQin, LISA, and TaiJi have the potential to outperform themselves through joint observation. To achieve this, it is desirable to practice joint data analysis in advance on simulated data that encodes the intrinsic correlation among the signals found in different detectors that operate simultaneously. In this paper, we introduce GWSpace, a package that can simulate the joint detection data from TianQin, LISA, and TaiJi. The software is not a groundbreaking work that starts from scratch. Rather, we use as many open-source resources as possible, tailoring them to the needs of simulating the multi-mission science data and putting everything into a ready-to-go and easy-to-use package. We shall describe the main components, the construction, and a few examples of application of the package. A common coordinate system, namely the Solar System Barycenter (SSB) coordinate system, is utilized to calculate spacecraft orbits for all three missions. The paper also provides a brief derivation of the detection process and outlines the general waveform of sources detectable by these detectors.

gr-qc↗

arXiv2025

On restricted-type Strichartz estimates and the applications

We establish a rigorous framework for the Zakharov system on waveguide manifolds $\mathbb{R}^m \times \mathbb{T}^n$ ($m,n\geq 1$), which models the nonlinear coupling between optical and acoustic modes in confined geometries such as optical fibers. Our analysis reveals that the sharp \textit{shell-type Strichartz estimate} for $\mathbb{R}^2 \times \mathbb{T}$ is globally valid in time and exhibits no derivative loss via the measure estimate of semi-algebraic sets, unlike the periodic case studied in \cite{MR4665720}. In addition, we demonstrate that such an estimate fails on the product space $\mathbb{R} \times \mathbb{T}^2$ by constructing a counter-example. Moreover, we derive analogues of these shell-type estimates in other dimensions, both in the waveguide and Euclidean settings. As a direct application, we establish, for the first time, a local well-posedness theory for the partially periodic Zakharov system. To summarize, we compare shell-type Strichartz estimates in different settings (the Euclidean, the periodic, and the waveguide). Numerical verification on $\mathbb{R}^2\times\mathbb{T}$ reveals a uniform $L^4$-spacetime bound, while $\mathbb{R}\times\mathbb{T}^2$ exhibits sublinear growth, quantitatively confirming the theoretical dichotomy between geometries with different dimensional confinement. These findings advance the understanding of dispersive effects in hybrid geometries and provide mathematical foundations for efficient waveguide design and signal transmission. Finally, for the Euclidean case, we establish well-posedness theory for supercritical nonlinear Schrödinger equation (NLS) with \textit{strip-type} frequency-restricted initial data, revealing a trade-off between dispersion and confinement, which is of independent mathematical interest. This provides a deterministic analogue to random data theory of NLS.

math.AP↗

arXiv2025

Quantum Parameter Estimation for Detectors in Constantly Accelerated Motion

We analyze quantum parameter estimation by studying the dynamics of the quantum Fisher information (QFI) for two classes of parameters, acceleration and initial-state weight, in an Unruh-DeWitt detector undergoing four distinct noninertial motions: linear, cusped, catenary, and circular trajectories respectively. We assume that the detector is initialized in a pure superposition state with a weight parameter $θ$ characterizing the probability of the detector occupying each state. Our results reveal that, over long evolution times, the QFI for the acceleration parameter converges to a nonnegative asymptotic value that depends sensitively on the trajectory, whereas the QFI for the weight parameter decays to zero as the system thermalizes. Importantly, for sufficiently large accelerations, one can attain the optimal precision in estimating the acceleration parameter within a finite interaction time, eliminating the need for infinitely long measurements. Comparing trajectories, we find that for small accelerations (relative to the detector's energy gap), linear motion yields the highest QFI for $θ$, while for large accelerations, circular motion becomes optimal for estimating $θ$. By contrast, circular motion offers the best precision for estimating acceleration itself in both the small- and large-acceleration regimes (the latter only at very long times). These contrasting behaviors of QFI across trajectories suggest a novel metrological protocol for inferring the underlying noninertial motion of a quantum probe.

quant-ph↗

arXiv2025

Can Large Language Models Simulate Human Responses? A Case Study of Stated Preference Experiments in the Context of Heating-related Choices

Stated preference (SP) surveys are a key method to research how individuals make trade-offs in hypothetical, also futuristic, scenarios. In energy context this includes key decarbonisation enablement contexts, such as low-carbon technologies, distributed renewable energy generation, and demand-side response [1,2]. However, they tend to be costly, time-consuming, and can be affected by respondent fatigue and ethical constraints. Large language models (LLMs) have demonstrated remarkable capabilities in generating human-like textual responses, prompting growing interest in their application to survey research. This study investigates the use of LLMs to simulate consumer choices in energy-related SP surveys and explores their integration into data analysis workflows. A series of test scenarios were designed to systematically assess the simulation performance of several LLMs (LLaMA 3.1, Mistral, GPT-3.5 and DeepSeek-R1) at both individual and aggregated levels, considering contexts factors such as prompt design, in-context learning (ICL), chain-of-thought (CoT) reasoning, LLM types, integration with traditional choice models, and potential biases. Cloud-based LLMs do not consistently outperform smaller local models. In this study, the reasoning model DeepSeek-R1 achieves the highest average accuracy (77%) and outperforms non-reasoning LLMs in accuracy, factor identification, and choice distribution alignment. Across models, systematic biases are observed against the gas boiler and no-retrofit options, with a preference for more energy-efficient alternatives. The findings suggest that previous SP choices are the most effective input factor, while longer prompts with additional factors and varied formats can cause LLMs to lose focus, reducing accuracy.

cs.CL↗

arXiv2025

LAMBench: A Benchmark for Large Atomistic Models

Large Atomistic Models (LAMs) have undergone remarkable progress recently, emerging as universal or fundamental representations of the potential energy surface defined by the first-principles calculations of atomistic systems. However, our understanding of the extent to which these models achieve true universality, as well as their comparative performance across different models, remains limited. This gap is largely due to the lack of comprehensive benchmarks capable of evaluating the effectiveness of LAMs as approximations to the universal potential energy surface. In this study, we introduce LAMBench, a benchmarking system designed to evaluate LAMs in terms of their generalizability, adaptability, and applicability. These attributes are crucial for deploying LAMs as ready-to-use tools across a diverse array of scientific discovery contexts. We benchmark ten state-of-the-art LAMs released prior to August 1, 2025, using LAMBench. Our findings reveal a significant gap between the current LAMs and the ideal universal potential energy surface. They also highlight the need for incorporating cross-domain training data, supporting multi-fidelity modeling, and ensuring the models' conservativeness and differentiability. As a dynamic and extensible platform, LAMBench is intended to continuously evolve, thereby facilitating the development of robust and generalizable LAMs capable of significantly advancing scientific research. The LAMBench code is open-sourced at https://github.com/deepmodeling/lambench, and an interactive leaderboard is available at https://www.aissquare.com/openlam?tab=Benchmark.

physics.comp-ph↗

arXiv2025

Stochastic Black Start Resource Allocation to Enable Dynamic Formation of Networked Microgrids and DER-aided Restoration

Extended outages in distributed systems (DSs) dominated by distributed energy resources (DERs) require innovative strategies to efficiently and securely deploy black start (BS) resources. To address the need, this paper proposes a two-stage stochastic resource allocation method within synchronizing dynamic microgrids (MGs) for black start (SDMG-BS), enabling risk-averse and adaptive restoration across various scenarios while ensuring frequency security. Virtual synchronous generator (VSG)-controlled grid-forming inverters (GFMIs) equipped with primary frequency governors (PFGs) are modeled as BS resources. Their frequency response is characterized by three transient indices, which are deployed as frequency dynamic constraints on load pick-up events to ensure frequency stability during the BS process. SDMG-BS framework facilitates location-independent synchronization among restored MGs and with the transmission grid (TG) with the help of smart switches (SSWs). The model incorporates scenario-based stochastic programming to address multi-source uncertainties, including season-dependent operational conditions and unpredictable TG outage durations, ensuring a resilient allocation plan. The proposed approach is validated on a modified IEEE 123-node feeder with three study cases designed across sixteen uncertainty scenarios.

eess.SY↗

arXiv2025

HateClipSeg: A Segment-Level Annotated Dataset for Fine-Grained Hate Video Detection

Detecting hate speech in videos remains challenging due to the complexity of multimodal content and the lack of fine-grained annotations in existing datasets. We present HateClipSeg, a large-scale multimodal dataset with both video-level and segment-level annotations, comprising over 11,714 segments labeled as Normal or across five Offensive categories: Hateful, Insulting, Sexual, Violence, Self-Harm, along with explicit target victim labels. Our three-stage annotation process yields high inter-annotator agreement (Krippendorff's alpha = 0.817). We propose three tasks to benchmark performance: (1) Trimmed Hateful Video Classification, (2) Temporal Hateful Video Localization, and (3) Online Hateful Video Classification. Results highlight substantial gaps in current models, emphasizing the need for more sophisticated multimodal and temporally aware approaches. The HateClipSeg dataset are publicly available at https://github.com/Social-AI-Studio/HateClipSeg.git.

cs.CV↗