arXiv subjects

Rui Yin

Publications and source records attributed to Rui Yin.

At least 19 recordsLinked to original sources

arXiv2026

Beyond Over-Refusal: Defending Indirect Prompt Injection via Latent Instruction Manifolds

Large Language Models (LLMs) have been integrated into complex ecosystems (e.g., Code Agents), while Indirect Prompt Injection (IPI) attacks have emerged as critical barriers to their safe deployment. Attackers exploit LLMs' indistinguishability between "instructions" and "data" to manipulate LLMs via maliciously injected instructions. Existing defenses, however, face an intractable safety-utility trade-off: most guardrails either incur high latency or suffer from severe over-refusal. In this paper, we first demonstrate that LLMs can separate instruction from data intrinsically with both theoretical and empirical evidence. Inspired by this insight, we propose AEGIS (Adaptive Ensemble Guard for Injection Shielding). AEGIS extracts instruction-sensitive projectors to identify malicious instructions and leverages a Unified Multi-Layer Consensus mechanism that aggregates topologically distinct signals across the network depth. Empirical evaluations show that AEGIS achieves remarkable detection performance against both heuristic and optimization-based attacks compared to baselines, highlighting its potential to mitigate IPI. Code is available at https://github.com/xaddwell/AEGIS

cs.CR↗

arXiv2026

The Boy Who Cried Wolf: Adversarial Misclassification of Safe Inputs as Unsafe in Multimodal Guardrails

Multimodal guard models have emerged as critical safety components for screening content in vision-language systems. While adversarial research has extensively studied jailbreaking attacks that produce false negatives, the inverse threat of inducing false positives on benign inputs remains unexplored. We introduce Unsafe Induction Attacks, where adversaries distribute imperceptibly perturbed safe images that trigger guard models to reject legitimate user requests, causing a "Boy Who Cried Wolf" effect that degrades service availability and erodes trust. This reveals an availability failure mode in deployed safety filters. To realize this threat under diverse user prompts, we propose Unsafe Semantic Distillation (USD), which aligns adversarial perturbations with distributional representations of unsafe content rather than prompt-specific instances. Evaluated on four state-of-the-art guard models across realistic user simulation scenarios, USD achieves 84% attack success rates, outperforming existing methods and exposing fundamental vulnerabilities in current multimodal safety architectures.

cs.CR↗

arXiv2026

LLM-as-a-Judge in Healthcare: A Scoping Analysis of Applications, Methods, and Human Alignment

Large language models (LLMs) are increasingly deployed across healthcare applications, including clinical documentation, diagnostic reasoning, medicine recommendation, and medical education. Their outputs are largely unstructured clinical text, which is difficult to reliably evaluate at scale. LLM-as-a-Judge, in which an LLM evaluates another system's output against task-specific criteria, offers a scalable alternative and is increasingly used in clinical evaluation, yet its validity in healthcare remains underexamined. Existing reviews focus on general-purpose LLM evaluation or on risk framework, rather than systematically characterizing how LLM-as-a-Judge is applied in healthcare and how well their judgments align with human experts. We therefore conduct a PRISMA-guided comprehensive review of LLM-as-a-Judge applications in healthcare, searching five databases for studies published between January 2023 and February 2026. After screening 541 records, 134 studies meet the eligibility and are coded by health scenario, judge configuration, technical approach, and validation design. LLM-as-a-Judge is concentrated in clinical decision support, clinical natural language processing (NLP), medical knowledge and question answering (QA), and medical communication. OpenAI models are the most frequently used judges, and prompt engineering appears in nearly all studies, with ensemble, multi-agent, and retrieval-augmented designs as common extensions. Among studies reporting human validation, LLM judges often show moderate to strong alignment with expert judgments, although reliability varies substantially across tasks. Overall, this review positions LLM-as-a-Judge as a promising framework for scalable healthcare AI evaluation, while emphasizing that its clinical value depends on model design and rigorous validation.

cs.CY↗

arXiv2026

Compiling Activation Steering into Weights via Null-Space Constraints for Stealthy Backdoors

Safety-aligned large language models (LLMs) are increasingly deployed in real-world pipelines, yet this deployment also enlarges the supply-chain attack surface: adversaries can distribute backdoored checkpoints that behave normally under standard evaluation but jailbreak when a hidden trigger is present. Recent post-hoc weight-editing methods offer an efficient approach to injecting such backdoors by directly modifying model weights to map a trigger to an attacker-specified response. However, existing methods typically optimize a token-level mapping that forces an affirmative prefix (e.g., ``Sure''), which does not guarantee sustained harmful output -- the model may begin with apparent agreement yet revert to safety-aligned refusal within a few decoding steps. We address this reliability gap by shifting the backdoor objective from surface tokens to internal representations. We extract a steering vector that captures the difference between compliant and refusal behaviors, and compile it into a persistent weight modification that activates only when the trigger is present. To preserve stealthiness and benign utility, we impose a null-space constraint so that the injected edit remains dormant on clean inputs. The method is efficient, requiring only a small set of examples and admitting a closed-form solution. Across multiple safety-aligned LLMs and jailbreak benchmarks, our method achieves high triggered attack success while maintaining non-triggered safety and general utility.

cs.CR↗

arXiv2026

Gastric-X: A Multimodal Multi-Phase Benchmark Dataset for Advancing Vision-Language Models in Gastric Cancer Analysis

Recent vision-language models (VLMs) have shown strong generalization and multimodal reasoning abilities in natural domains. However, their application to medical diagnosis remains limited by the lack of comprehensive and structured datasets that capture real clinical workflows. To advance the development of VLMs for clinical applications, particularly in gastric cancer, we introduce Gastric-X, a large-scale multimodal benchmark for gastric cancer analysis providing 1.7K cases. Each case in Gastric-X includes paired resting and dynamic CT scans, endoscopic image, a set of structured biochemical indicators, expert-authored diagnostic notes, and bounding box annotations of tumor regions, reflecting realistic clinical conditions. We systematically examine the capability of recent VLMs on five core tasks: Visual Question Answering (VQA), report generation, cross-modal retrieval, disease classification, and lesion localization. These tasks simulate critical stages of clinical workflow, from visual understanding and reasoning to multimodal decision support. Through this evaluation, we aim not only to assess model performance but also to probe the nature of VLM understanding: Can current VLMs meaningfully correlate biochemical signals with spatial tumor features and textual reports? We envision Gastric-X as a step toward aligning machine intelligence with the cognitive and evidential reasoning processes of physicians, and as a resource to inspire the development of next-generation medical VLMs.

cs.CV↗

arXiv2026

Stability as a Liability:Systematic Breakdown of Linguistic Structure in LLMs

Training stability is typically regarded as a prerequisite for reliable optimization in large language models. In this work, we analyze how stabilizing training dynamics affects the induced generation distribution. We show that under standard maximum likelihood training, stable parameter trajectories lead stationary solutions to approximately minimize the forward KL divergence to the empirical distribution, while implicitly reducing generative entropy. As a consequence, the learned model can concentrate probability mass on a limited subset of empirical modes, exhibiting systematic degeneration despite smooth loss convergence. We empirically validate this effect using a controlled feedback-based training framework that stabilizes internal generation statistics, observing consistent low-entropy outputs and repetitive behavior across architectures and random seeds. It indicates that optimization stability and generative expressivity are not inherently aligned, and that stability alone is an insufficient indicator of generative quality.

cs.AI↗

arXiv2025

Learning Patient-Specific Disease Dynamics with Latent Flow Matching for Longitudinal Imaging Generation

Understanding disease progression is a central clinical challenge with direct implications for early diagnosis and personalized treatment. While recent generative approaches have attempted to model progression, key mismatches remain: disease dynamics are inherently continuous and monotonic, yet latent representations are often scattered, lacking semantic structure, and diffusion-based models disrupt continuity with random denoising process. In this work, we propose to treat the disease dynamic as a velocity field and leverage Flow Matching (FM) to align the temporal evolution of patient data. Unlike prior methods, it captures the intrinsic dynamic of disease, making the progression more interpretable. However, a key challenge remains: in latent space, Auto-Encoders (AEs) do not guarantee alignment across patients or correlation with clinical-severity indicators (e.g., age and disease conditions). To address this, we propose to learn patient-specific latent alignment, which enforces patient trajectories to lie along a specific axis, with magnitude increasing monotonically with disease severity. This leads to a consistent and semantically meaningful latent space. Together, we present $\Delta$-LFM, a framework for modeling patient-specific latent progression with flow matching. Across three longitudinal MRI benchmarks, $\Delta$-LFM demonstrates strong empirical performance and, more importantly, offers a new framework for interpreting and visualizing disease dynamics.

cs.CV↗

arXiv2025

SenWave: A Fine-Grained Multi-Language Sentiment Analysis Dataset Sourced from COVID-19 Tweets

The global impact of the COVID-19 pandemic has highlighted the need for a comprehensive understanding of public sentiment and reactions. Despite the availability of numerous public datasets on COVID-19, some reaching volumes of up to 100 billion data points, challenges persist regarding the availability of labeled data and the presence of coarse-grained or inappropriate sentiment labels. In this paper, we introduce SenWave, a novel fine-grained multi-language sentiment analysis dataset specifically designed for analyzing COVID-19 tweets, featuring ten sentiment categories across five languages. The dataset comprises 10,000 annotated tweets each in English and Arabic, along with 30,000 translated tweets in Spanish, French, and Italian, derived from English tweets. Additionally, it includes over 105 million unlabeled tweets collected during various COVID-19 waves. To enable accurate fine-grained sentiment classification, we fine-tuned pre-trained transformer-based language models using the labeled tweets. Our study provides an in-depth analysis of the evolving emotional landscape across languages, countries, and topics, revealing significant insights over time. Furthermore, we assess the compatibility of our dataset with ChatGPT, demonstrating its robustness and versatility in various applications. Our dataset and accompanying code are publicly accessible on the repository\footnote{https://github.com/gitdevqiang/SenWave}. We anticipate that this work will foster further exploration into fine-grained sentiment analysis for complex events within the NLP community, promoting more nuanced understanding and research innovations.

cs.CL↗

arXiv2025

Talk Before You Retrieve: Agent-Led Discussions for Better RAG in Medical QA

Medical question answering (QA) is a reasoning-intensive task that remains challenging for large language models (LLMs) due to hallucinations and outdated domain knowledge. Retrieval-Augmented Generation (RAG) provides a promising post-training solution by leveraging external knowledge. However, existing medical RAG systems suffer from two key limitations: (1) a lack of modeling for human-like reasoning behaviors during information retrieval, and (2) reliance on suboptimal medical corpora, which often results in the retrieval of irrelevant or noisy snippets. To overcome these challenges, we propose Discuss-RAG, a plug-and-play module designed to enhance the medical QA RAG system through collaborative agent-based reasoning. Our method introduces a summarizer agent that orchestrates a team of medical experts to emulate multi-turn brainstorming, thereby improving the relevance of retrieved content. Additionally, a decision-making agent evaluates the retrieved snippets before their final integration. Experimental results on four benchmark medical QA datasets show that Discuss-RAG consistently outperforms MedRAG, especially significantly improving answer accuracy by up to 16.67% on BioASQ and 12.20% on PubMedQA. The code is available at: https://github.com/LLM-VLM-GSL/Discuss-RAG.

cs.CL↗

arXiv2025

High-Dimensional Encoding Computational Imaging

High-dimensional imaging technology has demonstrated significant research value across diverse fields, including environmental monitoring, agricultural inspection, and biomedical imaging, through integrating spatial (X*Y), spectral, and polarization detection functionalities. Here, we report a High-Dimensional encoding computational imaging technique, utilizing 4 high-dimensional encoders (HDE1-4) and a high-dimensional neural network (HDNN) to reconstruct 80 high-dimensional images of the target. The system efficiently acquires spectral-polarization information, spanning a wavelength range of 400-800 nm at intervals of 20 nm, obtaining 20 spectral datasets. Each dataset contains images captured at 4 polarization angles (0{\deg}, 45{\deg}, 90{\deg}, and -45{\deg}), and the image resolution can reach up to 1280 * 960 pixels. Achieving a reconstruction ratio 1:20. Experimental validation confirms that the spectral reconstruction error consistently remains below 0.14%. Extensive high-dimensional imaging experiments were conducted under indoor and outdoor conditions, showing the system's significant adaptability and robustness in various environments. Compared to traditional imaging devices, such as hyperspectral cameras that could only acquire spectral information, while polarization cameras are limited to polarization imaging, this integrated system successfully overcomes these technological constraints, providing an innovative and efficient solution for high-dimensional optical sensing applications.

physics.optics↗

arXiv2024

Sonicmesh: Enhancing 3D Human Mesh Reconstruction in Vision-Impaired Environments With Acoustic Signals

3D Human Mesh Reconstruction (HMR) from 2D RGB images faces challenges in environments with poor lighting, privacy concerns, or occlusions. These weaknesses of RGB imaging can be complemented by acoustic signals, which are widely available, easy to deploy, and capable of penetrating obstacles. However, no existing methods effectively combine acoustic signals with RGB data for robust 3D HMR. The primary challenges include the low-resolution images generated by acoustic signals and the lack of dedicated processing backbones. We introduce SonicMesh, a novel approach combining acoustic signals with RGB images to reconstruct 3D human mesh. To address the challenges of low resolution and the absence of dedicated processing backbones in images generated by acoustic signals, we modify an existing method, HRNet, for effective feature extraction. We also integrate a universal feature embedding technique to enhance the precision of cross-dimensional feature alignment, enabling SonicMesh to achieve high accuracy. Experimental results demonstrate that SonicMesh accurately reconstructs 3D human mesh in challenging environments such as occlusions, non-line-of-sight scenarios, and poor lighting.

cs.CV↗

arXiv2024

BiDense: Binarization for Dense Prediction

Dense prediction is a critical task in computer vision. However, previous methods often require extensive computational resources, which hinders their real-world application. In this paper, we propose BiDense, a generalized binary neural network (BNN) designed for efficient and accurate dense prediction tasks. BiDense incorporates two key techniques: the Distribution-adaptive Binarizer (DAB) and the Channel-adaptive Full-precision Bypass (CFB). The DAB adaptively calculates thresholds and scaling factors for binarization, effectively retaining more information within BNNs. Meanwhile, the CFB facilitates full-precision bypassing for binary convolutional layers undergoing various channel size transformations, which enhances the propagation of real-valued signals and minimizes information loss. By leveraging these techniques, BiDense preserves more real-valued information, enabling more accurate and detailed dense predictions in BNNs. Extensive experiments demonstrate that our framework achieves performance levels comparable to full-precision models while significantly reducing memory usage and computational costs.

cs.CV↗

arXiv2024

A novel and efficient parameter estimation of the Lognormal-Rician turbulence model based on k-Nearest Neighbor and data generation method

In this paper, we propose a novel and efficient parameter estimator based on $k$-Nearest Neighbor ($k$NN) and data generation method for the Lognormal-Rician turbulence channel. The Kolmogorov-Smirnov (KS) goodness-of-fit statistical tools are employed to investigate the validity of $k$NN approximation under different channel conditions and it is shown that the choice of $k$ plays a significant role in the approximation accuracy. We present several numerical results to illustrate that solving the constructed objective function can provide a reasonable estimate for the actual values. The accuracy of the proposed estimator is investigated in terms of the mean square error. The simulation results show that increasing the number of generation samples by two orders of magnitude does not lead to a significant improvement in estimation performance when solving the optimization problem by the gradient descent algorithm. However, the estimation performance under the genetic algorithm (GA) approximates to that of the saddlepoint approximation and expectation-maximization estimators. Therefore, combined with the GA, we demonstrate that the proposed estimator achieves the best tradeoff between the computation complexity and the accuracy.

eess.SP↗

arXiv2024

SRPose: Two-view Relative Pose Estimation with Sparse Keypoints

Two-view pose estimation is essential for map-free visual relocalization and object pose tracking tasks. However, traditional matching methods suffer from time-consuming robust estimators, while deep learning-based pose regressors only cater to camera-to-world pose estimation, lacking generalizability to different image sizes and camera intrinsics. In this paper, we propose SRPose, a sparse keypoint-based framework for two-view relative pose estimation in camera-to-world and object-to-camera scenarios. SRPose consists of a sparse keypoint detector, an intrinsic-calibration position encoder, and promptable prior knowledge-guided attention layers. Given two RGB images of a fixed scene or a moving object, SRPose estimates the relative camera or 6D object pose transformation. Extensive experiments demonstrate that SRPose achieves competitive or superior performance compared to state-of-the-art methods in terms of accuracy and speed, showing generalizability to both scenarios. It is robust to different image sizes and camera intrinsics, and can be deployed with low computing resources.

cs.CV↗

arXiv2024

Domain Game: Disentangle Anatomical Feature for Single Domain Generalized Segmentation

Single domain generalization aims to address the challenge of out-of-distribution generalization problem with only one source domain available. Feature distanglement is a classic solution to this purpose, where the extracted task-related feature is presumed to be resilient to domain shift. However, the absence of references from other domains in a single-domain scenario poses significant uncertainty in feature disentanglement (ill-posedness). In this paper, we propose a new framework, named \textit{Domain Game}, to perform better feature distangling for medical image segmentation, based on the observation that diagnostic relevant features are more sensitive to geometric transformations, whilist domain-specific features probably will remain invariant to such operations. In domain game, a set of randomly transformed images derived from a singular source image is strategically encoded into two separate feature sets to represent diagnostic features and domain-specific features, respectively, and we apply forces to pull or repel them in the feature space, accordingly. Results from cross-site test domain evaluation showcase approximately an ~11.8% performance boost in prostate segmentation and around ~10.5% in brain tumor segmentation compared to the second-best method.

cs.CV↗

arXiv2023

Optical Switching Data Center Networks: Understanding Techniques and Challenges

Relying on the flexible-access interconnects to the scalable storage and compute resources, data centers deliver critical communications connectivity among numerous servers to support the housed applications and services. To provide the high-speeds and long-distance communications, the data centers have turned to fiber interconnections. With the stringently increased traffic volume, the data centers are then expected to further deploy the optical switches into the systems infrastructure to implement the full optical switching. This paper first summarizes the topologies and traffic characteristics in data centers and analyzes the reasons and importance of moving to optical switching. Recent techniques related to the optical switching, and main challenges limiting the practical deployments of optical switches in data centers are also summarized and reported.

cs.NI↗

arXiv2023

New Results for the Pointing Errors Model in Two Asymptotic Cases

Several precise and computationally efficient results for pointing errors models in two asymptotic cases are derived in this paper. The normalized mean-squared error (NMSE) performance metric is employed to quantify the accuracy of different models. For the case that the beam width is relatively larger than the detection aperture, we propose the three kinds of models that have the form of $c_1\exp(-c_2r^2) $.It is shown that the modified intensity uniform model not only achieves a comparable accuracy with the best linearized model, but also is expressed in an elegant mathematical way when compared to the traditional Farid model. This indicates that the modified intensity uniform model is preferable in the performance analysis of free space optical (FSO) systems considering the effects of the pointing errors. By analogizing the beam spot with a point in the case that beam width is smaller than the detection aperture, the solution of the pointing errors model is transformed to a smooth function approximation problem, and we find that a more accurate approximation can be achieved by the proposed point approximation model when compared to the model that is induced from the Vasylyev model in some scenarios.

cs.IT↗

arXiv2022

Chaos bound of charged particles around phantom AdS black hole

In this paper, we investigated the chaos of the phantom AdS black hole in near-horizon regions and at a certain distance from the black hole, where the Lyapunov exponent was calculated by the Jacobian determinant. We found that the angular momenta of charged particles nearby the black hole affect the position of the equilibrium orbit as well as the Lyapunov exponent. When the charge is large enough and the angular momenta take particular values, the bound is violated at a certain distance from the event horizon. Concurrently, we compared the differences by analyzing tables and figures after taking different values of $\eta$, where $\eta$ is a constant indicating the nature of the electromagnetic field. When $\eta=1$, it corresponds to the Reissner-Nordstr$\ddot{o}$m-AdS black hole (RN-AdS black hole) and it represents the phantom AdS black hole when $\eta=-1$. In this way, we can draw the following conclusions. Under different values of $\eta$, the variation trends of curves related to $\lambda^{2}-\kappa^{2}$ could be different, where it can be judged whether the constraint is violated by determining if $\lambda^{2}-\kappa^{2}$ is greater than zero. In addition, the corresponding numerical values of angular momenta violating the bound in the phantom AdS black hole are much smaller than the case of RN-AdS black hole. But cases in these different black holes have the same rules to violate the bound. The large absolute values of charge and $\Lambda$ are more likely to violate the bound.

gr-qc↗