Efficient Jailbreak Mitigation Using Semantic Linear Classification in a Multi-Staged Pipeline

Prompt injection and jailbreaking attacks pose persistent security challenges to large language model (LLM)-based systems. We present an efficient and systematically evaluated defense architecture that mitigates these threats through a lightweight, multi-stage pipeline. Its core component is a semantic filter based on text normalization, TF-IDF representations, and a Linear SVM classifier. Despite its simplicity, this module achieves 93.4% accuracy and 96.5% specificity on held-out data, substantially reducing attack throughput while incurring negligible computational overhead. Building on this efficient foundation, the full pipeline integrates complementary detection and mitigation mechanisms that operate at successive stages, providing strong robustness with minimal latency. In comparative experiments, our SVM-based configuration improves overall accuracy from 35.1% to 93.4% while reducing average time to completion from approximately 450s to 47s, yielding over 10 times lower latency than ShieldGemma. These results demonstrate that the proposed design simultaneously advances defensive precision and efficiency, addressing a core limitation of current model-based moderators. Evaluation across a curated corpus of over 30,000 labeled prompts, including benign, jailbreak, and application-layer injections, confirms that staged, resource-efficient defenses can robustly secure modern LLM-driven applications.

LADLE-MM: Limited Annotation based Detector with Learned Ensembles for Multimodal Misinformation

With the rise of easily accessible tools for generating and manipulating multimedia content, realistic synthetic alterations to digital media have become a widespread threat, often involving manipulations across multiple modalities simultaneously. Recently, such techniques have been increasingly employed to distort narratives of important events and to spread misinformation on social media, prompting the development of misinformation detectors. In the context of misinformation conveyed through image-text pairs, several detection methods have been proposed. However, these approaches typically rely on computationally intensive architectures or require large amounts of annotated data. In this work we introduce LADLE-MM: Limited Annotation based Detector with Learned Ensembles for Multimodal Misinformation, a model-soup initialized multimodal misinformation detector designed to operate under a limited annotation setup and constrained training resources. LADLE-MM is composed of two unimodal branches and a third multimodal one that enhances image and text representations with additional multimodal embeddings extracted from BLIP, serving as fixed reference space. Despite using 60.3% fewer trainable parameters than previous state-of-the-art models, LADLE-MM achieves competitive performance on both binary and multi-label classification tasks on the DGM4 benchmark, outperforming existing methods when trained without grounding annotations. Moreover, when evaluated on the VERITE dataset, LADLE-MM outperforms current state-of-the-art approaches that utilize more complex architectures involving Large Vision-Language-Models, demonstrating the effective generalization ability in an open-set setting and strong robustness to unimodal bias.

Why Text Prevails: Vision May Undermine Multimodal Medical Decision Making

With the rapid progress of large language models (LLMs), advanced multimodal large language models (MLLMs) have demonstrated impressive zero-shot capabilities on vision-language tasks. In the biomedical domain, however, even state-of-the-art MLLMs struggle with basic Medical Decision Making (MDM) tasks. We investigate this limitation using two challenging datasets: (1) three-stage Alzheimer's disease (AD) classification (normal, mild cognitive impairment, dementia), where category differences are visually subtle, and (2) MIMIC-CXR chest radiograph classification with 14 non-mutually exclusive conditions. Our empirical study shows that text-only reasoning consistently outperforms vision-only or vision-text settings, with multimodal inputs often performing worse than text alone. To mitigate this, we explore three strategies: (1) in-context learning with reason-annotated exemplars, (2) vision captioning followed by text-only inference, and (3) few-shot fine-tuning of the vision tower with classification supervision. These findings reveal that current MLLMs lack grounded visual understanding and point to promising directions for improving multimodal decision making in healthcare.

R-GenIMA: Integrating Neuroimaging and Genetics with Interpretable Multimodal AI for Alzheimer's Disease Progression

Early detection of Alzheimer's disease (AD) requires models capable of integrating macro-scale neuroanatomical alterations with micro-scale genetic susceptibility, yet existing multimodal approaches struggle to align these heterogeneous signals. We introduce R-GenIMA, an interpretable multimodal large language model that couples a novel ROI-wise vision transformer with genetic prompting to jointly model structural MRI and single nucleotide polymorphisms (SNPs) variations. By representing each anatomically parcellated brain region as a visual token and encoding SNP profiles as structured text, the framework enables cross-modal attention that links regional atrophy patterns to underlying genetic factors. Applied to the ADNI cohort, R-GenIMA achieves state-of-the-art performance in four-way classification across normal cognition (NC), subjective memory concerns (SMC), mild cognitive impairment (MCI), and AD. Beyond predictive accuracy, the model yields biologically meaningful explanations by identifying stage-specific brain regions and gene signatures, as well as coherent ROI-Gene association patterns across the disease continuum. Attention-based attribution revealed genes consistently enriched for established GWAS-supported AD risk loci, including APOE, BIN1, CLU, and RBFOX1. Stage-resolved neuroanatomical signatures identified shared vulnerability hubs across disease stages alongside stage-specific patterns: striatal involvement in subjective decline, frontotemporal engagement during prodromal impairment, and consolidated multimodal network disruption in AD. These results demonstrate that interpretable multimodal AI can synthesize imaging and genetics to reveal mechanistic insights, providing a foundation for clinically deployable tools that enable earlier risk stratification and inform precision therapeutic strategies in Alzheimer's disease.

OmniMER: Indonesian Multimodal Emotion Recognition via Auxiliary-Enhanced LLM Adaptation

Indonesian, spoken by over 200 million people, remains underserved in multimodal emotion recognition research despite its dominant presence on Southeast Asian social media platforms. We introduce IndoMER, the first multimodal emotion recognition benchmark for Indonesian, comprising 1,944 video segments from 203 speakers with temporally aligned text, audio, and visual annotations across seven emotion categories. The dataset exhibits realistic challenges including cross-modal inconsistency and long-tailed class distributions shaped by Indonesian cultural communication norms. To address these challenges, we propose OmniMER, a multimodal adaptation framework built upon Qwen2.5-Omni that enhances emotion recognition through three auxiliary modality-specific perception tasks: emotion keyword extraction for text, facial expression analysis for video, and prosody analysis for audio. These auxiliary tasks help the model identify emotion-relevant cues in each modality before fusion, reducing reliance on spurious correlations in low-resource settings. Experiments on IndoMER show that OmniMER achieves 0.582 Macro-F1 on sentiment classification and 0.454 on emotion recognition, outperforming the base model by 7.6 and 22.1 absolute points respectively. Cross-lingual evaluation on the Chinese CH-SIMS dataset further demonstrates the generalizability of the proposed framework. The dataset and code are publicly available. https://github.com/yanxm01/INDOMER

The Effect of Negation on CLIP in Medical Imaging: Limitations of Contrastive Language-Image Pretraining

Large vision-language models like CLIP are increasingly used in medical imaging tasks due to their ability to align images and text without the need for extensive labeled data. This makes them particularly useful for applications like image retrieval, report generation, and classification in clinical settings. A potential issue to this approach is that CLIP-based models often under perform when interpreting negated phrases, which is especially problematic in the context of medical diagnosing. In this study, we evaluate the Stanford AIMI CheXagent model on its ability to correctly retrieve chest X-ray images using prompts with and without negation. The goal of this project is to understand where this model fails and then use it as a base model to improve its retrieval accuracy by fine tuning methods outlined in previous work. Results from this study show improvement in handling of negation in the CLIP model with a slight decrease in accuracy of positive prompt evaluation. Alongside retrieval accuracy, we examined internal model behavior through token attribution, t-SNE projection, and attention-head ablation to better characterize how each fine tuning approach reshaped the text encoders representation of negated clinical language. Through this work, we hope to better understand the internal behavior of CLIP and improve its handling of negation using clinically relevant language for improving its reliability in medical AI devices.

Auxiliary Descriptive Knowledge for Few-Shot Adaptation of Vision-Language Model

Despite the impressive zero-shot capabilities of Vision-Language Models (VLMs), they often struggle in downstream tasks with distribution shifts from the pre-training data. Few-Shot Adaptation (FSA-VLM) has emerged as a key solution, typically using Parameter-Efficient Fine-Tuning (PEFT) to adapt models with minimal data. However, these PEFT methods are constrained by their reliance on fixed, handcrafted prompts, which are often insufficient to understand the semantics of classes. While some studies have proposed leveraging image-induced prompts to provide additional clues for classification, they introduce prohibitive computational overhead at inference. Therefore, we introduce Auxiliary Descriptive Knowledge (ADK), a novel framework that efficiently enriches text representations without compromising efficiency. ADK first leverages a Large Language Model to generate a rich set of descriptive prompts for each class offline. These pre-computed features are then deployed in two ways: (1) as Compositional Knowledge, an averaged representation that provides rich semantics, especially beneficial when class names are ambiguous or unfamiliar to the VLM; and (2) as Instance-Specific Knowledge, where a lightweight, non-parametric attention mechanism dynamically selects the most relevant descriptions for a given image. This approach provides two additional types of knowledge alongside the handcrafted prompt, thereby facilitating category distinction across various domains. Also, ADK acts as a parameter-free, plug-and-play component that enhances existing PEFT methods. Extensive experiments demonstrate that ADK consistently boosts the performance of multiple PEFT baselines, setting a new state-of-the-art across various scenarios.

SLIM: Semantic-based Low-bitrate Image compression for Machines by leveraging diffusion

In recent years, the demand of image compression models for machine vision has increased dramatically. However, the training frameworks of image compression still focus on the vision of human, maintaining the excessive perceptual details, thus have limitations in optimally reducing the bits per pixel in the case of performing machine vision tasks. In this paper, we propose Semantic-based Low-bitrate Image compression for Machines by leveraging diffusion, termed SLIM. This is a new effective training framework of image compression for machine vision, using a pretrained latent diffusion model.The compressor model of our method focuses only on the Region-of-Interest (RoI) areas for machine vision in the image latent, to compress it compactly. Then the pretrained Unet model enhances the decompressed latent, utilizing a RoI-focused text caption which containing semantic information of the image. Therefore, SLIM is able to focus on RoI areas of the image without any guide mask at the inference stage, achieving low bitrate when compressing. And SLIM is also able to enhance a decompressed latent by denoising steps, so the final reconstructed image from the enhanced latent can be optimized for the machine vision task while still containing perceptual details for human vision. Experimental results show that SLIM achieves a higher classification accuracy in the same bits per pixel condition, compared to conventional image compression models for machines.

InstructNet: A Novel Approach for Multi-Label Instruction Classification through Advanced Deep Learning

People use search engines for various topics and items, from daily essentials to more aspirational and specialized objects. Therefore, search engines have taken over as peoples preferred resource. The How To prefix has become familiar and widely used in various search styles to find solutions to particular problems. This search allows people to find sequential instructions by providing detailed guidelines to accomplish specific tasks. Categorizing instructional text is also essential for task-oriented learning and creating knowledge bases. This study uses the How To articles to determine the multi-label instruction category. We have brought this work with a dataset comprising 11,121 observations from wikiHow, where each record has multiple categories. To find out the multi-label category meticulously, we employ some transformer-based deep neural architectures, such as Generalized Autoregressive Pretraining for Language Understanding (XLNet), Bidirectional Encoder Representation from Transformers (BERT), etc. In our multi-label instruction classification process, we have reckoned our proposed architectures using accuracy and macro f1-score as the performance metrics. This thorough evaluation showed us much about our strategys strengths and drawbacks. Specifically, our implementation of the XLNet architecture has demonstrated unprecedented performance, achieving an accuracy of 97.30% and micro and macro average scores of 89.02% and 93%, a noteworthy accomplishment in multi-label classification. This high level of accuracy and macro average score is a testament to the effectiveness of the XLNet architecture in our proposed InstructNet approach. By employing a multi-level strategy in our evaluation process, we have gained a more comprehensive knowledge of the effectiveness of our proposed architectures and identified areas for forthcoming improvement and refinement.

TWLR: Text-Guided Weakly-Supervised Lesion Localization and Severity Regression for Explainable Diabetic Retinopathy Grading

Accurate medical image analysis can greatly assist clinical diagnosis, but its effectiveness relies on high-quality expert annotations Obtaining pixel-level labels for medical images, particularly fundus images, remains costly and time-consuming. Meanwhile, despite the success of deep learning in medical imaging, the lack of interpretability limits its clinical adoption. To address these challenges, we propose TWLR, a two-stage framework for interpretable diabetic retinopathy (DR) assessment. In the first stage, a vision-language model integrates domain-specific ophthalmological knowledge into text embeddings to jointly perform DR grading and lesion classification, effectively linking semantic medical concepts with visual features. The second stage introduces an iterative severity regression framework based on weakly-supervised semantic segmentation. Lesion saliency maps generated through iterative refinement direct a progressive inpainting mechanism that systematically eliminates pathological features, effectively downgrading disease severity toward healthier fundus appearances. Critically, this severity regression approach achieves dual benefits: accurate lesion localization without pixel-level supervision and providing an interpretable visualization of disease-to-healthy transformations. Experimental results on the FGADR, DDR, and a private dataset demonstrate that TWLR achieves competitive performance in both DR classification and lesion segmentation, offering a more explainable and annotation-efficient solution for automated retinal image analysis.