--We propose a novel mechanism for real-time (human-in-the-loop) feedback focused on false positive reduction to enhance anomaly detection models. It was designed for the lightweight deployment of a behavioral network anomaly detection model. This methodology is easily integrable to similar domains that require a premium on throughput while maintaining high precision. In this paper, we introduce Retrieval Augmented Anomaly Detection, a novel method taking inspiration from Retrieval Augmented Generation. Human annotated examples are sent to a vector store, which can modify model outputs on the very next processed batch for model inference. T o demonstrate the generalization of this technique, we benchmarked several different model architectures and multiple data modalities, including images, text, and graph-based data. I NTRODUCTION Cybersecurity artificial intelligence (AI) models designed for network intrusion threat detection require very high, but nuanced, model precision.

Effective out-of-distribution (OOD) detection is crucial for the safe deployment of machine learning models in real-world scenarios. However, recent work has shown that OOD detection methods are vulnerable to adversarial attacks, potentially leading to critical failures in high-stakes applications. This discovery has motivated work on robust OOD detection methods that are capable of maintaining performance under various attack settings. Prior approaches have made progress on this problem but face a number of limitations: often only exhibiting robustness to attacks on OOD data or failing to maintain strong clean performance. In this work, we adapt an existing robust classification framework, TRADES, extending it to the problem of robust OOD detection and discovering a novel objective function. Recognising the critical importance of a strong clean/robust trade-off for OOD detection, we introduce an additional loss term which boosts classification and detection performance. Our approach, called HALO (Helper-based AdversariaL OOD detection), surpasses existing methods and achieves state-of-the-art performance across a number of datasets and attack settings. Extensive experiments demonstrate an average AUROC improvement of 3.15 in clean settings and 7.07 under adversarial attacks when compared to the next best method. Furthermore, HALO exhibits resistance to transferred attacks, offers tuneable performance through hyperparameter selection, and is compatible with existing OOD detection frameworks out-of-the-box, leaving open the possibility of future performance gains. Code is available at: https://github.com/hugo0076/HALO

Recent advancements in Artificial Intelligence have led to the development of Multimodal Large Language Models (MLLMs). However, adapting these pre-trained models to dynamic data distributions and various tasks efficiently remains a challenge. Fine-tuning MLLMs for specific tasks often causes performance degradation in the model's prior knowledge domain, a problem known as 'Catastrophic Forgetting'. While this issue has been well-studied in the Continual Learning (CL) community, it presents new challenges for MLLMs. This review paper, the first of its kind in MLLM continual learning, presents an overview and analysis of 440 research papers in this area.The review is structured into four sections. First, it discusses the latest research on MLLMs, covering model innovations, benchmarks, and applications in various fields. Second, it categorizes and overviews the latest studies on continual learning, divided into three parts: non-large language models unimodal continual learning (Non-LLM Unimodal CL), non-large language models multimodal continual learning (Non-LLM Multimodal CL), and continual learning in large language models (CL in LLM). The third section provides a detailed analysis of the current state of MLLM continual learning research, including benchmark evaluations, architectural innovations, and a summary of theoretical and empirical studies.Finally, the paper discusses the challenges and future directions of continual learning in MLLMs, aiming to inspire future research and development in the field. This review connects the foundational concepts, theoretical insights, method innovations, and practical applications of continual learning for multimodal large models, providing a comprehensive understanding of the research progress and challenges in this field, aiming to inspire researchers in the field and promote the advancement of related technologies.

Large language models (LLMs) have become the backbone of modern natural language processing but pose privacy concerns about leaking sensitive training data. Membership inference attacks (MIAs), which aim to infer whether a sample is included in a model's training dataset, can serve as a foundation for broader privacy threats. Existing defenses designed for traditional classification models do not account for the sequential nature of text data. As a result, they either require significant computational resources or fail to effectively mitigate privacy risks in LLMs. In this work, we propose a lightweight yet effective empirical privacy defense for protecting training data of language modeling by leveraging the token-specific characteristics. By analyzing token dynamics during training, we propose a token selection strategy that categorizes tokens into hard tokens for learning and memorized tokens for unlearning. Subsequently, our training-phase defense optimizes a novel dual-purpose token-level loss to achieve a Pareto-optimal balance between utility and privacy. Extensive experiments demonstrate that our approach not only provides strong protection against MIAs but also improves language modeling performance by around 10\% across various LLM architectures and datasets compared to the baselines.

Cardiovascular diseases are a leading cause of death and disability worldwide. Electrocardiogram (ECG) recordings are critical for diagnosing and monitoring cardiac health, but obtaining large-scale annotated ECG datasets is labor-intensive and time-consuming. Recent ECG Self-Supervised Learning (eSSL) methods mitigate this by learning features without extensive labels but fail to capture fine-grained clinical semantics and require extensive task-specific fine-tuning. To address these challenges, we propose $\textbf{SuPreME}$, a $\textbf{Su}$pervised $\textbf{Pre}$-training framework for $\textbf{M}$ultimodal $\textbf{E}$CG representation learning. SuPreME applies Large Language Models (LLMs) to extract structured clinical entities from free-text ECG reports, filter out noise and irrelevant content, enhance clinical representation learning, and build a high-quality, fine-grained labeled dataset. By using text-based cardiac queries instead of traditional categorical labels, SuPreME enables zero-shot classification of unseen diseases without additional fine-tuning. We evaluate SuPreME on six downstream datasets covering 127 cardiac conditions, achieving superior zero-shot AUC performance over state-of-the-art eSSL and multimodal methods by over 1.96\%. Results demonstrate the effectiveness of SuPreME in leveraging structured, clinically relevant knowledge for high-quality ECG representations. All code and data will be released upon acceptance.

Machine learning systems trained on electronic health records (EHRs) increasingly guide treatment decisions, but their reliability depends on the critical assumption that patients follow the prescribed treatments recorded in EHRs. Using EHR data from 3,623 hypertension patients, we investigate how treatment non-adherence introduces implicit bias that can fundamentally distort both causal inference and predictive modeling. By extracting patient adherence information from clinical notes using a large language model, we identify 786 patients (21.7%) with medication non-adherence. We further uncover key demographic and clinical factors associated with non-adherence, as well as patient-reported reasons including side effects and difficulties obtaining refills. Our findings demonstrate that this implicit bias can not only reverse estimated treatment effects, but also degrade model performance by up to 5% while disproportionately affecting vulnerable populations by exacerbating disparities in decision outcomes and model error rates. This highlights the importance of accounting for treatment non-adherence in developing responsible and equitable clinical machine learning systems.

Peer review is a critical process for ensuring the integrity of published scientific research. Confidence in this process is predicated on the assumption that experts in the relevant domain give careful consideration to the merits of manuscripts which are submitted for publication. With the recent rapid advancements in large language models (LLMs), a new risk to the peer review process is that negligent reviewers will rely on LLMs to perform the often time consuming process of reviewing a paper. However, there is a lack of existing resources for benchmarking the detectability of AI text in the domain of peer review. To address this deficiency, we introduce a comprehensive dataset containing a total of 788,984 AI-written peer reviews paired with corresponding human reviews, covering 8 years of papers submitted to each of two leading AI research conferences (ICLR and NeurIPS). We use this new resource to evaluate the ability of 18 existing AI text detection algorithms to distinguish between peer reviews written by humans and different state-of-the-art LLMs. Motivated by the shortcomings of existing methods, we propose a new detection approach which surpasses existing methods in the identification of AI written peer reviews. Our work reveals the difficulty of identifying AI-generated text at the individual peer review level, highlighting the urgent need for new tools and methods to detect this unethical use of generative AI.

Decision-making in complex systems often relies on machine learning models, yet highly accurate models such as XGBoost and neural networks can obscure the reasoning behind their predictions. In operations research applications, understanding how a decision is made is often as crucial as the decision itself. Traditional interpretable models, such as decision trees and logistic regression, provide transparency but may struggle with datasets containing intricate feature interactions. However, complexity in decision-making stem from interactions that are only relevant within certain subsets of data. Within these subsets, feature interactions may be simplified, forming simple structures where simple interpretable models can perform effectively. We propose a bottom-up simple structure-identifying algorithm that partitions data into interpretable subgroups known as simple structure, where feature interactions are minimized, allowing simple models to be trained within each subgroup. We demonstrate the robustness of the algorithm on synthetic data and show that the decision boundaries derived from simple structures are more interpretable and aligned with the intuition of the domain than those learned from a global model. By improving both explainability and predictive accuracy, our approach provides a principled framework for decision support in applications where model transparency is essential.

Large-scale machine learning models deliver strong performance across a wide range of tasks but come with significant computational and resource constraints. To mitigate these challenges, local smaller models are often deployed alongside larger models, relying on routing and deferral mechanisms to offload complex tasks. However, existing approaches inadequately balance the capabilities of these models, often resulting in unnecessary deferrals or sub-optimal resource usage. In this work we introduce a novel loss function called Gatekeeper for calibrating smaller models in cascade setups. Our approach fine-tunes the smaller model to confidently handle tasks it can perform correctly while deferring complex tasks to the larger model. Moreover, it incorporates a mechanism for managing the trade-off between model performance and deferral accuracy, and is broadly applicable across various tasks and domains without any architectural changes. We evaluate our method on encoder-only, decoder-only, and encoder-decoder architectures. Experiments across image classification, language modeling, and vision-language tasks show that our approach substantially improves deferral performance.

Dementia is a progressive cognitive syndrome with Alzheimer's disease (AD) as the leading cause. Conversation-based AD detection offers a cost-effective alternative to clinical methods, as language dysfunction is an early biomarker of AD. However, most prior research has framed AD detection as a binary classification problem, limiting the ability to identify Mild Cognitive Impairment (MCI)-a crucial stage for early intervention. Also, studies primarily rely on single-language datasets, mainly in English, restricting cross-language generalizability. To address this gap, we make three key contributions. First, we introduce a novel, multilingual dataset for AD detection by unifying 16 publicly available dementia-related conversational datasets. This corpus spans English, Spanish, Chinese, and Greek and incorporates both audio and text data derived from a variety of cognitive assessment tasks. Second, we perform finer-grained classification, including MCI, and evaluate various classifiers using sparse and dense text representations. Third, we conduct experiments in monolingual and multilingual settings, finding that some languages benefit from multilingual training while others perform better independently. This study highlights the challenges in multilingual AD detection and enables future research on both language-specific approaches and techniques aimed at improving model generalization and robustness.