arXiv subjects

Yunfan Yang

Publications and source records attributed to Yunfan Yang.

At least 19 recordsLinked to original sources

arXiv2026

CSPO: Alleviating Reward Ambiguity for Structured Table-to-LaTeX Generation

Tables contain rich structured information, yet when stored as images their contents remain "locked" within pixels. Converting table images into LaTeX code enables faithful digitization and reuse, but current multimodal large language models (MLLMs) often fail to preserve structural, style, or content fidelity. Conventional post-training with reinforcement learning (RL) typically relies on a single aggregated reward, leading to reward ambiguity that conflates multiple behavioral aspects and hinders effective optimization. We propose Component-Specific Policy Optimization (CSPO), an RL framework that disentangles optimization across LaTeX tables components-structure, style, and content. In particular, CSPO assigns component-specific rewards and backpropagates each signal only through the tokens relevant to its component, alleviating reward ambiguity and enabling targeted component-wise optimization. To comprehensively assess performance, we introduce a set of hierarchical evaluation metrics. Extensive experiments demonstrate the effectiveness of CSPO, underscoring the importance of component-specific optimization for reliable structured generation. Our code is available at https://github.com/microsoft/CSPO.

cs.AI↗

arXiv2026

Breaking the mutual exclusivity between metallicity and ferroelectricity in a non-polar covalent semiconductor via orbital selective doping

The mutual exclusion of ferroelectricity and metallic conductivity is a long-standing tenet because itinerant electrons screen long-range Coulomb forces that stabilize the bulk polar order. Here, we break this paradigm by heavily doping a non-polar covalent semiconductor of cubic silicon carbide (3C-SiC) with nitrogen. This introduces heavy electron doping, inducing metallicity and driving a structural transition from the non-polar F-43m to the polar R3m symmetry via the pseudo-Jahn-Teller effect. Remarkably, we provide direct, atomic-scale visualization of about 180° polarization reversal under an external voltage bias in a ferroelectric metal. The strongly directional character of antibonding orbitals occupied by conduction electrons prevents them from screening the local Si-C polarization, resulting in the coexistence of metallicity and ferroelectricity. Ferroelectric tunnel junctions demonstrate nonvolatile memory properties with a well-defined high-resistance state (HRS) and low-resistance state (LRS), an ultrahigh response speed (~50 ns), an ultralow operating voltage (1 V), an endurance exceeding 85927 cycles, and a projected retention time of 100 years. Our results provide a novel strategy for pioneering ferroelectricity in a metal, a new ferroelectric metal platform for exploring exotic properties, and a ferroelectric device with high performance that meets the requirements for low consumption and high-speed non-volatile devices.

cond-mat.mtrl-sci↗

arXiv2026

A plug-and-play generative framework for multi-satellite precipitation estimation

Reliable precipitation monitoring is essential for disaster risk reduction, water resources management, and agricultural decision-making. Multi-source satellite observations, particularly the combination of geostationary infrared and passive microwave measurements, have become a primary means of precipitation detection. Traditional multi-source satellite precipitation estimation methods remain computationally inefficient, and many deep learning methods lack the flexibility to incorporate new sensors without retraining the full model. Here we introduce PRISMA (Precipitation Inference from Satellite Modalities via generAtive modeling), a plug-and-play latent generative framework for multi-sensor precipitation estimation. PRISMA learns an unconditional precipitation prior from IMERG Final fields and constrains it through independently trained, sensor-specific conditional branches, allowing new observation sources to be incorporated without retraining the generative backbone. Applied to FY-4B AGRI infrared and GPM GMI microwave observations, PRISMA improves Critical Success Index by up to 40.3% and reduces root-mean-square error by 22.6% relative to infrared-only estimation within microwave swaths, while also improving probabilistic skill and maintaining an average inference time of about 37 s. Independent rain-gauge validation across China confirms consistent gains, and typhoon case studies show that microwave conditioning restores eyewall and spiral rainband structures, reducing storm-core mean absolute error by up to 42.3%. PRISMA thus provides an extensible and efficient framework for multi-sensor precipitation estimation.

physics.ao-ph↗

arXiv2026

StormDiT: A generative AI model bridges the 2-6 hour 'gray zone' in precipitation nowcasting

Accurate short-term warnings for extreme precipitation are critical for global disaster mitigation but are hindered by a persistent predictability barrier at the 2-6 hour horizon -- the "nowcasting gray zone." In this window, traditional observation-based extrapolation fails due to error accumulation, while numerical weather prediction is computationally too slow to resolve storm-scale dynamics. Recent generative AI approaches attempt to bridge this gap by decomposing precipitation into separate deterministic advection and stochastic diffusion components. However, this decomposition can sever fundamental causal links between entangled atmospheric processes, such as the dynamic initiation of convection triggered by boundary advection. Here we present StormDiT, a unified generative model that treats weather evolution as a holistic spatiotemporal problem, learning the coupled physics of the gray zone without human-imposed structural priors. Trained on a massive dataset of 7,720 precipitation events from China, our model achieves a breakthrough in long-horizon stability. On a heavy-rainfall test set, it maintains skillful prediction for strong convection ($\ge$ 35 dBZ) with a Critical Success Index (CSI) near 0.2 across the full 6-hour forecast at 6-minute resolution. Crucially, the model exhibits superior probabilistic calibration, accurately quantifying operational risks. On the public SEVIR benchmark, our unified paradigm more than doubles the state-of-the-art 1-hour performance for heavy rain and establishes the first robust baseline for 3-hour forecasting. Furthermore, interpretability analysis reveals that the model attends to non-local physical precursors, such as outflow boundaries, explicitly validating its emergent understanding of convective organization.

physics.ao-ph↗

arXiv2026

Nowcast3D: Reliable precipitation nowcasting via gray-box learning

Reliable nowcasting of extreme precipitation remains difficult because convective systems are strongly nonlinear, multiscale, and nonstationary in 3D. Radar is the backbone of nowcasting, yet existing methods struggle to predict extremes: physics-based extrapolation cannot capture growth and decay, deterministic learning tends to oversmooth and underestimate peaks, and purely generative models often lack physical consistency. Hybrid schemes help but are mostly limited to 2D composite reflectivity, collapsing the atmosphere into one layer and discarding vertical structure critical for height-dependent dynamics. We introduce Nowcast3D, a gray-box, fully 3D framework that works directly on volumetric radar reflectivity. The end-to-end model couples physically constrained neural operators (advection, local diffusion, and microphysics) with a conditional diffusion model to generate ensemble forecasts with quantified uncertainty. Trained on provincial-scale 3D volumes over a $10.24^\circ \times 10.24^\circ$ region and fine-tuned on a $2.56^\circ \times 2.56^\circ$ city region ($0.01^\circ \approx 1$ km), Nowcast3D provides near-real-time forecasts up to 3 h and outperforms competitive baselines in cross-region and temporal out-of-sample tests. It can also infer wind fields without labeled supervision, supporting physically plausible transport. In a nationwide blind evaluation by 160 meteorologists, Nowcast3D ranked first and was preferred in 57% of post-hoc assessments, surpassing the leading baseline (27%). These results highlight its reliability and operational value for extreme precipitation nowcasting.

cs.LG↗

arXiv2026

Global Parametric Gates for Multi-qubit Entanglement

We propose and experimentally demonstrate a global parametric gate that generates multi-qubit entangled states in a single step. By applying a parametric drive to a common qubit at precise detunings relative to computational qubits, we directly produce two-, three-, and four-qubit entanglement with state fidelities of 99.4\%\pm0.2\%, 93.4\%\pm0.3\%, and 91.4\%\pm0.3\%, respectively. This scheme enables efficient, reconfigurable control using only microwave drives and is compatible with fixed-frequency qubits. Error analyses indicate that infidelity stems primarily from decoherence and coherent control errors, with negligible contributions from static ZZ coupling and flux noise. Furthermore, simulations with state-of-the-art parameters predict this global gate can generate high-fidelity (99.70\%) entanglement in systems of up to six qubits.

quant-ph↗

arXiv2025

Secure Confidential Business Information When Sharing Machine Learning Models

Model-sharing offers significant business value by enabling firms with well-established Machine Learning (ML) models to monetize and share their models with others who lack the resources to develop ML models from scratch. However, concerns over data confidentiality remain a significant barrier to model-sharing adoption, as Confidential Property Inference (CPI) attacks can exploit shared ML models to uncover confidential properties of the model provider's private model training data. Existing defenses often assume that CPI attacks are non-adaptive to the specific ML model they are targeting. This assumption overlooks a key characteristic of real-world adversaries: their responsiveness, i.e., adversaries' ability to dynamically adjust their attack models based on the information of the target and its defenses. To overcome this limitation, we propose a novel defense method that explicitly accounts for the responsive nature of real-world adversaries via two methodological innovations: a novel Responsive CPI attack and an attack-defense arms race framework. The former emulates the responsive behaviors of adversaries in the real world, and the latter iteratively enhances both the target and attack models, ultimately producing a secure ML model that is robust against responsive CPI attacks. Furthermore, we propose and integrate a novel approximate strategy into our defense, which addresses a critical computational bottleneck of defense methods and improves defense efficiency. Through extensive empirical evaluations across various realistic model-sharing scenarios, we demonstrate that our method outperforms existing defenses by more effectively defending against CPI attacks, preserving ML model utility, and reducing computational overhead.

cs.CR↗

arXiv2025

AnyAttack: Towards Large-scale Self-supervised Adversarial Attacks on Vision-language Models

Due to their multimodal capabilities, Vision-Language Models (VLMs) have found numerous impactful applications in real-world scenarios. However, recent studies have revealed that VLMs are vulnerable to image-based adversarial attacks. Traditional targeted adversarial attacks require specific targets and labels, limiting their real-world impact.We present AnyAttack, a self-supervised framework that transcends the limitations of conventional attacks through a novel foundation model approach. By pre-training on the massive LAION-400M dataset without label supervision, AnyAttack achieves unprecedented flexibility - enabling any image to be transformed into an attack vector targeting any desired output across different VLMs.This approach fundamentally changes the threat landscape, making adversarial capabilities accessible at an unprecedented scale. Our extensive validation across five open-source VLMs (CLIP, BLIP, BLIP2, InstructBLIP, and MiniGPT-4) demonstrates AnyAttack's effectiveness across diverse multimodal tasks. Most concerning, AnyAttack seamlessly transfers to commercial systems including Google Gemini, Claude Sonnet, Microsoft Copilot and OpenAI GPT, revealing a systemic vulnerability requiring immediate attention.

cs.LG↗

arXiv2025

Mind with Eyes: from Language Reasoning to Multimodal Reasoning

Language models have recently advanced into the realm of reasoning, yet it is through multimodal reasoning that we can fully unlock the potential to achieve more comprehensive, human-like cognitive capabilities. This survey provides a systematic overview of the recent multimodal reasoning approaches, categorizing them into two levels: language-centric multimodal reasoning and collaborative multimodal reasoning. The former encompasses one-pass visual perception and active visual perception, where vision primarily serves a supporting role in language reasoning. The latter involves action generation and state update within reasoning process, enabling a more dynamic interaction between modalities. Furthermore, we analyze the technical evolution of these methods, discuss their inherent challenges, and introduce key benchmark tasks and evaluation metrics for assessing multimodal reasoning performance. Finally, we provide insights into future research directions from the following two perspectives: (i) from visual-language reasoning to omnimodal reasoning and (ii) from multimodal reasoning to multimodal agents. This survey aims to provide a structured overview that will inspire further advancements in multimodal reasoning research.

cs.CL↗

arXiv2025

Debiased Prompt Tuning in Vision-Language Model without Annotations

Prompt tuning of Vision-Language Models (VLMs) such as CLIP, has demonstrated the ability to rapidly adapt to various downstream tasks. However, recent studies indicate that tuned VLMs may suffer from the problem of spurious correlations, where the model relies on spurious features (e.g. background and gender) in the data. This may lead to the model having worse robustness in out-of-distribution data. Standard methods for eliminating spurious correlation typically require us to know the spurious attribute labels of each sample, which is hard in the real world. In this work, we explore improving the group robustness of prompt tuning in VLMs without relying on manual annotation of spurious features. We notice the zero - shot image recognition ability of VLMs and use this ability to identify spurious features, thus avoiding the cost of manual annotation. By leveraging pseudo-spurious attribute annotations, we further propose a method to automatically adjust the training weights of different groups. Extensive experiments show that our approach efficiently improves the worst-group accuracy on CelebA, Waterbirds, and MetaShift datasets, achieving the best robustness gap between the worst-group accuracy and the overall accuracy.

cs.CV↗

arXiv2024

Debiasing Vison-Language Models with Text-Only Training

Pre-trained vision-language models (VLMs), such as CLIP, have exhibited remarkable performance across various downstream tasks by aligning text and images in a unified embedding space. However, due to the imbalanced distribution of pre-trained datasets, CLIP suffers from the bias problem in real-world applications. Existing debiasing methods struggle to obtain sufficient image samples for minority groups and incur high costs for group labeling. To address the limitations, we propose a Text-Only Debiasing framework called TOD, leveraging a text-as-image training paradigm to mitigate visual biases. Specifically, this approach repurposes the text encoder to function as an image encoder, thereby eliminating the need for image data. Simultaneously, it utilizes a large language model (LLM) to generate a balanced text dataset, which is then used for prompt tuning. However, we observed that the model overfits to the text modality because label names, serving as supervision signals, appear explicitly in the texts. To address this issue, we further introduce a Multi-Target Prediction (MTP) task that motivates the model to focus on complex contexts and distinguish between target and biased information. Extensive experiments on the Waterbirds and CelebA datasets show that our method significantly improves group robustness, achieving state-of-the-art results among image-free methods and even competitive performance compared to image-supervised methods. Furthermore, the proposed method can be adapted to challenging scenarios with multiple or unknown bias attributes, demonstrating its strong generalization and robustness.

cs.CV↗

arXiv2024

Thermal Conductivity of Cubic Silicon Carbide Single Crystals Heavily Doped by Nitrogen

High-purity cubic silicon carbide possesses the second-highest thermal conductivity among large-scale crystals, surpassed only by diamond, making it crucial for practical applications of thermal management. Recent theoretical studies predict that heavy doping reduces the thermal conductivity of 3C-SiC via phonon-defect and phonon-electron scattering. However, experimental evidence has been limited. In this work, we report the thermal conductivity of heavily nitrogen doped 3C SiC single crystals, grown using the top-seeded solution growth method, measured via time domain thermoreflectance. Our results show that a significant reduction (up to 30%) in thermal conductivity is observed with nitrogen doping concentrations around 1020 cm-3. A comparison with theoretical calculations indicates less intensive scatterings are observed in the measured thermal conductivity. We speculate that the electron-phonon scattering may have a smaller impact than previously anticipated or the distribution of defects are nonuniform which leads to less intensive scatterings. These findings shed light on understanding the doping effects on thermal transport in semiconductors and support further exploration of 3C SiC for thermal management in electronics.

cond-mat.mtrl-sci↗

arXiv2023

Efficient Encoding of Graphics Primitives with Simplex-based Structures

Grid-based structures are commonly used to encode explicit features for graphics primitives such as images, signed distance functions (SDF), and neural radiance fields (NeRF) due to their simple implementation. However, in $n$-dimensional space, calculating the value of a sampled point requires interpolating the values of its $2^n$ neighboring vertices. The exponential scaling with dimension leads to significant computational overheads. To address this issue, we propose a simplex-based approach for encoding graphics primitives. The number of vertices in a simplex-based structure increases linearly with dimension, making it a more efficient and generalizable alternative to grid-based representations. Using the non-axis-aligned simplicial structure property, we derive and prove a coordinate transformation, simplicial subdivision, and barycentric interpolation scheme for efficient sampling, which resembles transformation procedures in the simplex noise algorithm. Finally, we use hash tables to store multiresolution features of all interest points in the simplicial grid, which are passed into a tiny fully connected neural network to parameterize graphics primitives. We implemented a detailed simplex-based structure encoding algorithm in C++ and CUDA using the methods outlined in our approach. In the 2D image fitting task, the proposed method is capable of fitting a giga-pixel image with 9.4% less time compared to the baseline method proposed by instant-ngp, while maintaining the same quality and compression rate. In the volumetric rendering setup, we observe a maximum 41.2% speedup when the samples are dense enough.

cs.CV↗

arXiv2023

Promoting Open-domain Dialogue Generation through Learning Pattern Information between Contexts and Responses

Recently, utilizing deep neural networks to build the opendomain dialogue models has become a hot topic. However, the responses generated by these models suffer from many problems such as responses not being contextualized and tend to generate generic responses that lack information content, damaging the user's experience seriously. Therefore, many studies try introducing more information into the dialogue models to make the generated responses more vivid and informative. Unlike them, this paper improves the quality of generated responses by learning the implicit pattern information between contexts and responses in the training samples. In this paper, we first build an open-domain dialogue model based on the pre-trained language model (i.e., GPT-2). And then, an improved scheduled sampling method is proposed for pre-trained models, by which the responses can be used to guide the response generation in the training phase while avoiding the exposure bias problem. More importantly, we design a response-aware mechanism for mining the implicit pattern information between contexts and responses so that the generated replies are more diverse and approximate to human replies. Finally, we evaluate the proposed model (RAD) on the Persona-Chat and DailyDialog datasets; and the experimental results show that our model outperforms the baselines on most automatic and manual metrics.

cs.CL↗

arXiv2023

Towards Black-box Adversarial Example Detection: A Data Reconstruction-based Method

Adversarial example detection is known to be an effective adversarial defense method. Black-box attack, which is a more realistic threat and has led to various black-box adversarial training-based defense methods, however, does not attract considerable attention in adversarial example detection. In this paper, we fill this gap by positioning the problem of black-box adversarial example detection (BAD). Data analysis under the introduced BAD settings demonstrates (1) the incapability of existing detectors in addressing the black-box scenario and (2) the potential of exploring BAD solutions from a data perspective. To tackle the BAD problem, we propose a data reconstruction-based adversarial example detection method. Specifically, we use variational auto-encoder (VAE) to capture both pixel and frequency representations of normal examples. Then we use reconstruction error to detect adversarial examples. Compared with existing detection methods, the proposed method achieves substantially better detection performance in BAD, which helps promote the deployment of adversarial example detection-based defense solutions in real-world models.

cs.CV↗

arXiv2023

High quality and wafer-scale cubic silicon carbide single crystals

Silicon carbide (SiC) is an important semiconductor material for fabricating power electronic devices that exhibit higher switch frequency, lower energy loss and substantial reduction both in size and weight in comparison with its Si-based counterparts1-4. Currently, most devices, such as metal-oxide-semiconductor field effect transistors, which are core devices used in electric vehicles, photovoltaic industry and other applications, are fabricated on a hexagonal polytype 4H-SiC because of its commercial availability5. Cubic silicon carbide (3C-SiC), the only cubic polytype, has a moderate band gap of 2.36 eV at room-temperature, but a superior mobility and thermal conduction than 4H-SiC4,6-11. Moreover, the much lower concentration of interfacial traps between insulating oxide gate and 3C-SiC helps fabricate reliable and long-life devices7-10,12-14. The growth of 3C-SiC crystals, however, has remained a challenge up to now despite of decades-long efforts by researchers because of its easy transformation into other polytypes during growth15-19, limiting the 3C-SiC based devices. Here, we report that 3C-SiC can be made thermodynamically favored from nucleation to growth on a 4H-SiC substrate by top-seeded solution growth technique(TSSG), beyond what's expected by classic nucleation theory. This enables the steady growth of quality and large sized 3C-SiC crystals (2~4-inch in diameter and 4.0~10.0 mm in thickness) sustainable. Our findings broaden the mechanism of hetero-seed crystal growth and provide a feasible route to mass production of 3C-SiC crystals,offering new opportunities to develop power electronic devices potentially with better performances than those based on 4H-SiC.

cond-mat.mtrl-sci↗

arXiv2022

ImageNet Pre-training also Transfers Non-Robustness

ImageNet pre-training has enabled state-of-the-art results on many tasks. In spite of its recognized contribution to generalization, we observed in this study that ImageNet pre-training also transfers adversarial non-robustness from pre-trained model into fine-tuned model in the downstream classification tasks. We first conducted experiments on various datasets and network backbones to uncover the adversarial non-robustness in fine-tuned model. Further analysis was conducted on examining the learned knowledge of fine-tuned model and standard model, and revealed that the reason leading to the non-robustness is the non-robust features transferred from ImageNet pre-trained model. Finally, we analyzed the preference for feature learning of the pre-trained model, explored the factors influencing robustness, and introduced a simple robust ImageNet pre-training solution. Our code is available at \url{https://github.com/jiamingzhang94/ImageNet-Pretraining-transfers-non-robustness}.

cs.CV↗

arXiv2019

Precise Clock Synchronization in the Readout Electronics of WCDA in LHAASO

The Water Cherenkov Detector Array (WCDA) is one of the key parts in the Large High Altitude Air Shower Observatory (LHAASO). In the WCDA, 3600 Photomultiplier Tubes (PMTs) and the Front End Electronics (FEEs) are scattered within a 90000 m2 area, while a time measurement resolution better than 0.5 ns is required in the readout electronics. To achieve such time measurement precision, high quality clock distribution and synchronization among the 400 FEEs (each FEE for 9 PMTs readout) is required. To simplify the electronics system architecture, data, commands, and clock are transmitted simultaneously through fibers over a 400-meter distance between FEEs and the Clock and Data Transfer Modules (CDTMs). In this article, we propose a new method based on the White Rabbit (WR) to achieve completely automatic clock phase alignment between different FEEs. The original WR is enhanced to overcome the clock delay fluctuations due to ambient temperature variations. This paper presents the general scheme, the design of prototype electronics, and initial test results. These indicate that a clock synchronization precision better than 50 ps is achieved over 1 km fibers, which is well beyond the application requirement.

physics.ins-det↗