arXiv subjects

Yue Wu

Publications and source records attributed to Yue Wu.

At least 163 records · Page 9Linked to original sources

arXiv2025

BackdoorMBTI: A Backdoor Learning Multimodal Benchmark Tool Kit for Backdoor Defense Evaluation

Over the past few years, the emergence of backdoor attacks has presented significant challenges to deep learning systems, allowing attackers to insert backdoors into neural networks. When data with a trigger is processed by a backdoor model, it can lead to mispredictions targeted by attackers, whereas normal data yields regular results. The scope of backdoor attacks is expanding beyond computer vision and encroaching into areas such as natural language processing and speech recognition. Nevertheless, existing backdoor defense methods are typically tailored to specific data modalities, restricting their application in multimodal contexts. While multimodal learning proves highly applicable in facial recognition, sentiment analysis, action recognition, visual question answering, the security of these models remains a crucial concern. Specifically, there are no existing backdoor benchmarks targeting multimodal applications or related tasks. In order to facilitate the research in multimodal backdoor, we introduce BackdoorMBTI, the first backdoor learning toolkit and benchmark designed for multimodal evaluation across three representative modalities from eleven commonly used datasets. BackdoorMBTI provides a systematic backdoor learning pipeline, encompassing data processing, data poisoning, backdoor training, and evaluation. The generated poison datasets and backdoor models enable detailed evaluation of backdoor defenses. Given the diversity of modalities, BackdoorMBTI facilitates systematic evaluation across different data types. Furthermore, BackdoorMBTI offers a standardized approach to handling practical factors in backdoor learning, such as issues related to data quality and erroneous labels. We anticipate that BackdoorMBTI will expedite future research in backdoor defense methods within a multimodal context. Code is available at https://github.com/SJTUHaiyangYu/BackdoorMBTI.

cs.CR↗

arXiv2025

Scattered Forest Search: Smarter Code Space Exploration with LLMs

We frame code generation as a black-box optimization problem within the code space and demonstrate how optimization-inspired techniques can enhance inference scaling. Based on this perspective, we propose SCATTERED FOREST SEARCH (SFS), a novel approach that improves solution diversity and better exploits feedback during evolutionary search. Our theoretical analysis illustrates how these methods help avoid local optima during optimization, leading to more efficient exploration. Extensive experiments on HumanEval, MBPP, APPS, CodeContests, and Leetcode reveal significant performance gains. For instance, our method achieves a pass@1 rate of 67.1% on HumanEval+ and 87.2% on HumanEval with GPT-3.5, marking improvements of 8.6% and 4.3% over the state-of-the-art, while also halving the iterations needed to find the correct solution. Furthermore, our approach scales more efficiently than existing search techniques, including tree search, line search, and repeated sampling.

cs.SE↗

arXiv2025

DeepSeek-R1 Outperforms Gemini 2.0 Pro, OpenAI o1, and o3-mini in Bilingual Complex Ophthalmology Reasoning

Purpose: To evaluate the accuracy and reasoning ability of DeepSeek-R1 and three other recently released large language models (LLMs) in bilingual complex ophthalmology cases. Methods: A total of 130 multiple-choice questions (MCQs) related to diagnosis (n = 39) and management (n = 91) were collected from the Chinese ophthalmology senior professional title examination and categorized into six topics. These MCQs were translated into English using DeepSeek-R1. The responses of DeepSeek-R1, Gemini 2.0 Pro, OpenAI o1 and o3-mini were generated under default configurations between February 15 and February 20, 2025. Accuracy was calculated as the proportion of correctly answered questions, with omissions and extra answers considered incorrect. Reasoning ability was evaluated through analyzing reasoning logic and the causes of reasoning error. Results: DeepSeek-R1 demonstrated the highest overall accuracy, achieving 0.862 in Chinese MCQs and 0.808 in English MCQs. Gemini 2.0 Pro, OpenAI o1, and OpenAI o3-mini attained accuracies of 0.715, 0.685, and 0.692 in Chinese MCQs (all P<0.001 compared with DeepSeek-R1), and 0.746 (P=0.115), 0.723 (P=0.027), and 0.577 (P<0.001) in English MCQs, respectively. DeepSeek-R1 achieved the highest accuracy across five topics in both Chinese and English MCQs. It also excelled in management questions conducted in Chinese (all P<0.05). Reasoning ability analysis showed that the four LLMs shared similar reasoning logic. Ignoring key positive history, ignoring key positive signs, misinterpretation medical data, and too aggressive were the most common causes of reasoning errors. Conclusion: DeepSeek-R1 demonstrated superior performance in bilingual complex ophthalmology reasoning tasks than three other state-of-the-art LLMs. While its clinical applicability remains challenging, it shows promise for supporting diagnosis and clinical decision-making.

cs.CL↗

arXiv2025

Instance-Dependent Regret Bounds for Learning Two-Player Zero-Sum Games with Bandit Feedback

No-regret self-play learning dynamics have become one of the premier ways to solve large-scale games in practice. Accelerating their convergence via improving the regret of the players over the naive $O(\sqrt{T})$ bound after $T$ rounds has been extensively studied in recent years, but almost all studies assume access to exact gradient feedback. We address the question of whether acceleration is possible under bandit feedback only and provide an affirmative answer for two-player zero-sum normal-form games. Specifically, we show that if both players apply the Tsallis-INF algorithm of Zimmert and Seldin (2018, arXiv:1807.07623), then their regret is at most $O(c_1 \log T + \sqrt{c_2 T})$, where $c_1$ and $c_2$ are game-dependent constants that characterize the difficulty of learning -- $c_1$ resembles the complexity of learning a stochastic multi-armed bandit instance and depends inversely on some gap measures, while $c_2$ can be much smaller than the number of actions when the Nash equilibria have a small support or are close to the boundary. In particular, for the case when a pure strategy Nash equilibrium exists, $c_2$ becomes zero, leading to an optimal instance-dependent regret bound as we show. We additionally prove that in this case, our algorithm also enjoys last-iterate convergence and can identify the pure strategy Nash equilibrium with near-optimal sample complexity.

cs.LG↗

arXiv2025

Micro Blossom: Accelerated Minimum-Weight Perfect Matching Decoding for Quantum Error Correction

Minimum-Weight Perfect Matching (MWPM) decoding is important to quantum error correction decoding because of its accuracy. However, many believe that it is difficult, if possible at all, to achieve the microsecond latency requirement posed by superconducting qubits. This work presents the first publicly known MWPM decoder, called Micro Blossom, that achieves sub-microsecond decoding latency. Micro Blossom employs a heterogeneous architecture that carefully partitions a state-of-the-art MWPM decoder between software and a programmable accelerator with parallel processing units, one of each vertex/edge of the decoding graph. On a surface code with code distance $d$ and a circuit-level noise model with physical error rate $p$, Micro Blossom's accelerator employs $O(d^3)$ parallel processing units to reduce the worst-case latency from $O(d^{12})$ to $O(d^9)$ and reduce the average latency from $O(p d^3+1)$ to $O(p^2 d^2+1)$ when $p \ll 1$. We report a prototype implementation of Micro Blossom using FPGA. Measured at $d=13$ and $p=0.1\%$, the prototype achieves an average decoding latency of $0.8 μs$ at a moderate clock frequency of $62 MHz$. Micro Blossom is the first publicly known hardware-accelerated exact MWPM decoder, and the decoding latency of $0.8 μs$ is 8 times shorter than the best latency of MWPM decoder implementations reported in the literature.

cs.AR↗

arXiv2025

Fundus2Globe: Generative AI-Driven 3D Digital Twins for Personalized Myopia Management

Myopia, projected to affect 50% population globally by 2050, is a leading cause of vision loss. Eyes with pathological myopia exhibit distinctive shape distributions, which are closely linked to the progression of vision-threatening complications. Recent understanding of eye-shape-based biomarkers requires magnetic resonance imaging (MRI), however, it is costly and unrealistic in routine ophthalmology clinics. We present Fundus2Globe, the first AI framework that synthesizes patient-specific 3D eye globes from ubiquitous 2D color fundus photographs (CFPs) and routine metadata (axial length, spherical equivalent), bypassing MRI dependency. By integrating a 3D morphable eye model (encoding biomechanical shape priors) with a latent diffusion model, our approach achieves submillimeter accuracy in reconstructing posterior ocular anatomy efficiently. Fundus2Globe uniquely quantifies how vision-threatening lesions (e.g., staphylomas) in CFPs correlate with MRI-validated 3D shape abnormalities, enabling clinicians to simulate posterior segment changes in response to refractive shifts. External validation demonstrates its robust generation performance, ensuring fairness across underrepresented groups. By transforming 2D fundus imaging into 3D digital replicas of ocular structures, Fundus2Globe is a gateway for precision ophthalmology, laying the foundation for AI-driven, personalized myopia management.

eess.IV↗

arXiv2025

Structure-aware Domain Knowledge Injection for Large Language Models

This paper introduces a pioneering methodology, termed StructTuning, to efficiently transform foundation Large Language Models (LLMs) into domain specialists. It significantly reduces the training corpus needs to a mere 5% while achieving an impressive 100% of traditional knowledge injection performance. Motivated by structured human education, we propose a novel two-stage strategy for knowledge injection and alignment: Structure-aware Continual Pre-Training (SCPT) and Structure-aware Supervised Fine-Tuning (SSFT). In the SCPT phase, we automatically extract the domain knowledge taxonomy and reorganize the training corpora, enabling LLMs to effectively link textual segments to targeted knowledge points within the taxonomy. In the SSFT phase, we explicitly prompt models to elucidate the underlying knowledge structure in their outputs, leveraging the structured domain insight to address practical problems. Our ultimate method was extensively evaluated across model architectures and scales on LongBench and MMedBench datasets, demonstrating superior performance against other knowledge injection methods. We also explored our method's scalability across different training corpus sizes, laying the foundation to enhance domain-specific LLMs with better data utilization.

cs.CL↗

arXiv2025

MATH-Perturb: Benchmarking LLMs' Math Reasoning Abilities against Hard Perturbations

Large language models have demonstrated impressive performance on challenging mathematical reasoning tasks, which has triggered the discussion of whether the performance is achieved by true reasoning capability or memorization. To investigate this question, prior work has constructed mathematical benchmarks when questions undergo simple perturbations -- modifications that still preserve the underlying reasoning patterns of the solutions. However, no work has explored hard perturbations, which fundamentally change the nature of the problem so that the original solution steps do not apply. To bridge the gap, we construct MATH-P-Simple and MATH-P-Hard via simple perturbation and hard perturbation, respectively. Each consists of 279 perturbed math problems derived from level-5 (hardest) problems in the MATH dataset (Hendrycksmath et. al., 2021). We observe significant performance drops on MATH-P-Hard across various models, including o1-mini (-16.49%) and gemini-2.0-flash-thinking (-12.9%). We also raise concerns about a novel form of memorization where models blindly apply learned problem-solving skills without assessing their applicability to modified contexts. This issue is amplified when using original problems for in-context learning. We call for research efforts to address this challenge, which is critical for developing more robust and reliable reasoning models.

cs.LG↗

arXiv2025

Finite difference alternative WENO schemes with Riemann invariant-based local characteristic decompositions for compressible Euler equations

The weighted essentially non-oscillatory (WENO) schemes are widely used for hyperbolic conservation laws due to the ability to resolve discontinuities and maintain high-order accuracy in smooth regions at the same time. For hyperbolic systems, the WENO procedure is usually performed on local characteristic variables that are obtained by local characteristic decompositions to avoid oscillation near shocks. However, such decompositions are often computationally expensive. In this paper, we study a Riemann invariant-based local characteristic decomposition for the compressible Euler equations that reduces the cost. We apply the WENO procedure to the local characteristic fields of the Riemann invariants, where the eigenmatrix is sparse and thus the computational cost can be reduced. It is difficult to obtain the cell averages of Riemann invariants from those of the conserved variables due to the nonlinear relation between them, so we only focus on the finite difference alternative WENO versions. The efficiency and non-oscillatory property of the proposed schemes are well demonstrated by our numerical results.

math.NA↗

arXiv2025

Self-Regulation and Requesting Interventions

Human intelligence involves metacognitive abilities like self-regulation, recognizing limitations, and seeking assistance only when needed. While LLM Agents excel in many domains, they often lack this awareness. Overconfident agents risk catastrophic failures, while those that seek help excessively hinder efficiency. A key challenge is enabling agents with a limited intervention budget $C$ is to decide when to request assistance. In this paper, we propose an offline framework that trains a "helper" policy to request interventions, such as more powerful models or test-time compute, by combining LLM-based process reward models (PRMs) with tabular reinforcement learning. Using state transitions collected offline, we score optimal intervention timing with PRMs and train the helper model on these labeled trajectories. This offline approach significantly reduces costly intervention calls during training. Furthermore, the integration of PRMs with tabular RL enhances robustness to off-policy data while avoiding the inefficiencies of deep RL. We empirically find that our method delivers optimal helper behavior.

cs.LG↗

arXiv2025

Explicit positivity preserving numerical method for linear stochastic volatility models driven by $α$-stable process

In this paper, we introduce a linear stochastic volatility model driven by $α$-stable processes, which admits a unique positive solution. To preserve positivity, we modify the classical forward Euler-Maruyama scheme and analyze its numerical properties. The scheme achieves a strong convergence order of $1/α$. Numerical simulations are presented at the end to verify theoretical results.

math.PR↗

arXiv2025

Projected Langevin Monte Carlo algorithms in non-convex and super-linear setting

It is of significant interest in many applications to sample from a high-dimensional target distribution $π$ with the density $π(\text{d} x) \propto e^{-U(x)} (\text{d} x) $, based on the temporal discretization of the Langevin stochastic differential equations (SDEs). In this paper, we propose an explicit projected Langevin Monte Carlo (PLMC) algorithm with non-convex potential $U$ and super-linear gradient of $U$ and investigate the non-asymptotic analysis of its sampling error in total variation distance. Equipped with time-independent regularity estimates for the associated Kolmogorov equation, we derive the non-asymptotic bounds on the total variation distance between the target distribution of the Langevin SDEs and the law induced by the PLMC scheme with order $\mathcal{O}(d^{\max\{3γ/2 , 2γ-1 \}} h |\ln h|)$, where $d$ is the dimension of the target distribution and $γ\geq 1$ characterizes the growth of the gradient of $U$. In addition, if the gradient of $U$ is globally Lipschitz continuous, an improved convergence order of $\mathcal{O}(d^{3/2} h)$ for the classical Langevin Monte Carlo (LMC) scheme is derived with a refinement of the proof based on Malliavin calculus techniques. To achieve a given precision $ε$, the smallest number of iterations of the PLMC algorithm is proved to be of order ${\mathcal{O}}\big(\tfrac{d^{\max\{3γ/2 , 2γ-1 \}}}ε \ \cdot \ln (\tfrac{d}ε) \cdot \ln (\tfrac{1}ε) \big)$. In particular, the classical Langevin Monte Carlo (LMC) scheme with the non-convex potential $U$ and the globally Lipschitz gradient of $U$ can be guaranteed by order ${\mathcal{O}}\big(\tfrac{d^{3/2}}ε \cdot \ln (\tfrac{1}ε) \big)$. Numerical experiments are provided to confirm the theoretical findings.

math.NA↗

arXiv2025

Randomised Euler-Maruyama method for SDEs with Hölder continuous drift coefficient

In this paper, we examine the performance of randomised Euler-Maruyama (EM) method for additive time-inhomogeneous SDEs with an irregular drift. In particular, the drift is assumed to be $α$-Hölder continuous in time and bounded $β$-Hölder continuous in space with $α,β\in (0,1]$. The strong order of convergence of the randomised EM in $L^p$-norm is shown to be $1/2+(α\wedge (β/2))-ε$ for an arbitrary $ε\in (0,1/2)$, higher than the one of standard EM, which is $α\wedge (1/2+β/2-ε)$. The proofs highly rely on the stochastic sewing lemma, where we also provide an alternative proof when handling time irregularity for a comparison.

math.PR↗

arXiv2025

T2V-CompBench: A Comprehensive Benchmark for Compositional Text-to-video Generation

Text-to-video (T2V) generative models have advanced significantly, yet their ability to compose different objects, attributes, actions, and motions into a video remains unexplored. Previous text-to-video benchmarks also neglect this important ability for evaluation. In this work, we conduct the first systematic study on compositional text-to-video generation. We propose T2V-CompBench, the first benchmark tailored for compositional text-to-video generation. T2V-CompBench encompasses diverse aspects of compositionality, including consistent attribute binding, dynamic attribute binding, spatial relationships, motion binding, action binding, object interactions, and generative numeracy. We further carefully design evaluation metrics of multimodal large language model (MLLM)-based, detection-based, and tracking-based metrics, which can better reflect the compositional text-to-video generation quality of seven proposed categories with 1400 text prompts. The effectiveness of the proposed metrics is verified by correlation with human evaluations. We also benchmark various text-to-video generative models and conduct in-depth analysis across different models and various compositional categories. We find that compositional text-to-video generation is highly challenging for current models, and we hope our attempt could shed light on future research in this direction.

cs.CV↗

arXiv2025

Order-one explicit approximations of random periodic solutions of semi-linear SDEs with multiplicative noise

This paper is devoted to order-one explicit approximations of random periodic solutions to multiplicative noise driven stochastic differential equations (SDEs) with non-globally Lipschitz coefficients. The existence of the random periodic solution is demonstrated as the limit of the pull-back of the discretized SDE. A novel approach is introduced to analyze mean-square error bounds of the proposed scheme that does not depend on a prior high-order moment bounds of the numerical approximations. Under mild assumptions, the proposed scheme is proved to achieve an expected order-one mean square convergence in the infinite time horizon. Numerical examples are finally provided to verify the theoretical results.

math.PR↗

arXiv2024

Energy stable neural network for gradient flow equations

We propose an energy stable network (EStable-Net) for solving gradient flow equations. The EStable-Net enables decreasing of a discrete energy along the neural network, which is consistent with the property of the gradient flow equation. The architecture of the neural network EStable-Net is based on the block network structure (Autoflow) in which output of each block can be interpreted as an intermediate state of the evolution process of the equation, and the energy stable property is incorporated in each block, which is easily generalized to include other physical and/or numerical properties. Our EStable-Net is a supervised learning network approach for solving evolution equations which does not depend on the convergence of time step goes to 0, and can be applied generally even when only data is available but the equation is unknown. We also propose a training strategy for supervised learning that employs data of the evolution stages with different nature. The EStable-Net is validated by numerical experimental results based on the Allen-Cahn equation and the Cahn-Hilliard equation in two dimensions.

cs.LG↗

arXiv2024

OCTCube-M: A 3D multimodal optical coherence tomography foundation model for retinal and systemic diseases with cross-cohort and cross-device validation

We present OCTCube-M, a 3D OCT-based multi-modal foundation model for jointly analyzing OCT and en face images. OCTCube-M first developed OCTCube, a 3D foundation model pre-trained on 26,685 3D OCT volumes encompassing 1.62 million 2D OCT images. It then exploits a novel multi-modal contrastive learning framework COEP to integrate other retinal imaging modalities, such as fundus autofluorescence and infrared retinal imaging, into OCTCube, efficiently extending it into multi-modal foundation models. OCTCube achieves best performance on predicting 8 retinal diseases, demonstrating strong generalizability on cross-cohort, cross-device and cross-modality prediction. OCTCube can also predict cross-organ nodule malignancy (CT) and low cardiac ejection fraction as well as systemic diseases, such as diabetes and hypertension, revealing its wide applicability beyond retinal diseases. We further develop OCTCube-IR using COEP with 26,685 OCT and IR image pairs. OCTCube-IR can accurately retrieve between OCT and IR images, allowing joint analysis between 3D and 2D retinal imaging modalities. Finally, we trained a tri-modal foundation model OCTCube-EF from 4 million 2D OCT images and 400K en face retinal images. OCTCube-EF attains the best performance on predicting the growth rate of geographic atrophy (GA) across datasets collected from 6 multi-center global trials conducted in 23 countries. This improvement is statistically equivalent to running a clinical trial with more than double the size of the original study. Our analysis based on another retrospective case study reveals OCTCube-EF's ability to avoid false positive Phase-III results according to its accurate treatment effect estimation on the Phase-II results. In sum, OCTCube-M is a 3D multi-modal foundation model framework that integrates OCT and other retinal imaging modalities revealing substantial diagnostic and prognostic benefits.

eess.IV↗

arXiv2024

Semantically Grounded QFormer for Efficient Vision Language Understanding

General purpose Vision Language Models (VLMs) have received tremendous interest in recent years, owing to their ability to learn rich vision-language correlations as well as their broad zero-shot competencies. One immensely popular line of work utilizes frozen unimodal models, by bridging vision representations to language using a trainable module called the QFormer. However, this method relies heavily on large-scale multimodal pretraining with huge computational overheads. To that end, we propose a more efficient framework for QFormer-based vision-language alignment. Our key idea relies on the observation that QFormer latents correspond more strongly to the frozen LLM's intermediate latent space. Consequently, instead of using QFormer latents as inputs to the LLM, we alter the framework by using the latents to directly condition the LLM latent space for image-to-text generation. We demonstrate the effectiveness of our approach against existing baselines in improving the efficiency of vision-language pretraining.

cs.CV↗