Steering the behavior of Large Language Models (LLMs) remains a challenge, particularly in engineering applications where precision and reliability are critical. While fine-tuning and prompting methods can modify model behavior, they lack the dynamic and exact control necessary for engineering applications. Inference-time intervention techniques provide a promising alternative, allowing targeted adjustments to LLM outputs. In this work, we demonstrate how interventions enable fine-grained control for automating the usually time-intensive requirement verification process in Model-Based Systems Engineering (MBSE). Using two early-stage Capella SysML models of space missions with associated requirements, we apply the intervened LLMs to reason over a graph representation of the model to determine whether a requirement is fulfilled. Our method achieves robust and reliable outputs, significantly improving over both a baseline model and a fine-tuning approach. By identifying and modifying as few as one to three specialised attention heads, we can significantly change the model's behavior. When combined with self-consistency, this allows us to achieve perfect precision on our holdout test set.

Long-term test-time adaptation (TTA) is a challenging task due to error accumulation. Recent approaches tackle this issue by actively labeling a small proportion of samples in each batch, yet the annotation burden quickly grows as the batch number increases. In this paper, we investigate how to achieve effortless active labeling so that a maximum of one sample is selected for annotation in each batch. First, we annotate the most valuable sample in each batch based on the single-step optimization perspective in the TTA context. In this scenario, the samples that border between the source- and target-domain data distributions are considered the most feasible for the model to learn in one iteration. Then, we introduce an efficient strategy to identify these samples using feature perturbation. Second, we discover that the gradient magnitudes produced by the annotated and unannotated samples have significant variations. Therefore, we propose balancing their impact on model optimization using two dynamic weights. Extensive experiments on the popular ImageNet-C, -R, -K, -A and PACS databases demonstrate that our approach consistently outperforms state-of-the-art methods with significantly lower annotation costs.

As deep neural networks(DNN) become increasingly prevalent, particularly in high-stakes areas such as autonomous driving and healthcare, the ability to detect incorrect predictions of models and intervene accordingly becomes crucial for safety. In this work, we investigate the detection of misclassified inputs for image classification models from the lens of adversarial perturbation: we propose to use robust radius (a.k.a. input-space margin) as a confidence metric and design two efficient estimation algorithms, RR-BS and RR-Fast, for misclassification detection. Furthermore, we design a training method called Radius Aware Training (RAT) to boost models' ability to identify mistakes. Extensive experiments show our method could achieve up to 29.3% reduction on AURC and 21.62% reduction in FPR@95TPR, compared with previous methods.

As large language models (LLMs) are increasingly used for factual question-answering, it becomes more important for LLMs to have the capability to communicate the likelihood that their answer is correct. For these verbalized expressions of uncertainty to be meaningful, they should reflect the error rates at the expressed level of confidence. However, when prompted to express confidence, the error rates of current LLMs are inconsistent with their communicated confidences, highlighting the need for uncertainty quantification methods. Many prior methods calculate lexical uncertainty, estimating a model's confidence in the specific string it generated. In some cases, however, it may be more useful to estimate semantic uncertainty, or the model's confidence in the answer regardless of how it is verbalized. We propose a simple procedure, uncertainty distillation, to teach an LLM to verbalize calibrated semantic confidences. Using held-out data to map initial uncertainty estimates to meaningful probabilities, we create examples annotated with verbalized probabilities for supervised fine-tuning. We demonstrate our method yields verbalized confidences that correlate with observed error rates with a small fine-tuned language model as well as with larger instruction-tuned models, and find that our semantic uncertainty correlates well with lexical uncertainty on short answers.

Sepsis is a life-threatening syndrome with high morbidity and mortality in hospitals. Early prediction of sepsis plays a crucial role in facilitating early interventions for septic patients. However, early sepsis prediction systems with uncertainty quantification and adaptive learning are scarce. This paper proposes Sepsyn-OLCP, a novel online learning algorithm for early sepsis prediction by integrating conformal prediction for uncertainty quantification and Bayesian bandits for adaptive decision-making. By combining the robustness of Bayesian models with the statistical uncertainty guarantees of conformal prediction methodologies, this algorithm delivers accurate and trustworthy predictions, addressing the critical need for reliable and adaptive systems in high-stakes healthcare applications such as early sepsis prediction. We evaluate the performance of Sepsyn-OLCP in terms of regret in stochastic bandit setting, the area under the receiver operating characteristic curve (AUROC), and F-measure. Our results show that Sepsyn-OLCP outperforms existing individual models, increasing AUROC of a neural network from 0.64 to 0.73 without retraining and high computational costs. And the model selection policy converges to the optimal strategy in the long run. We propose a novel reinforcement learning-based framework integrated with conformal prediction techniques to provide uncertainty quantification for early sepsis prediction. The proposed methodology delivers accurate and trustworthy predictions, addressing a critical need in high-stakes healthcare applications like early sepsis prediction.

Named Entity Recognition (NER) is a critical component of Natural Language Processing (NLP) for extracting structured information from unstructured text. However, for low-resource languages like Catalan, the performance of NER systems often suffers due to the lack of high-quality annotated datasets. This paper introduces NERCat, a fine-tuned version of the GLiNER[1] model, designed to improve NER performance specifically for Catalan text. We used a dataset of manually annotated Catalan television transcriptions to train and fine-tune the model, focusing on domains such as politics, sports, and culture. The evaluation results show significant improvements in precision, recall, and F1-score, particularly for underrepresented named entity categories such as Law, Product, and Facility. This study demonstrates the effectiveness of domain-specific fine-tuning in low-resource languages and highlights the potential for enhancing Catalan NLP applications through manual annotation and high-quality datasets.

False arrhythmia alarms in intensive care units (ICUs) are a significant challenge, contributing to alarm fatigue and potentially compromising patient safety. Ventricular tachycardia (VT) alarms are particularly difficult to detect accurately due to their complex nature. This paper presents a machine learning approach to reduce false VT alarms using the VTaC dataset, a benchmark dataset of annotated VT alarms from ICU monitors. We extract time-domain and frequency-domain features from waveform data, preprocess the data, and train deep learning models to classify true and false VT alarms. Our results demonstrate high performance, with ROC-AUC scores exceeding 0.96 across various training configurations. This work highlights the potential of machine learning to improve the accuracy of VT alarm detection in clinical settings.

Multimodal large language models (MLLMs) have made remarkable strides in cross-modal comprehension and generation tasks. However, they remain vulnerable to jailbreak attacks, where crafted perturbations bypass security guardrails and elicit harmful outputs. In this paper, we present the first adversarial training (AT) paradigm tailored to defend against jailbreak attacks during the MLLM training phase. Extending traditional AT to this domain poses two critical challenges: efficiently tuning massive parameters and ensuring robustness against attacks across multiple modalities. To address these challenges, we introduce Projection Layer Against Adversarial Training (ProEAT), an end-to-end AT framework. ProEAT incorporates a projector-based adversarial training architecture that efficiently handles large-scale parameters while maintaining computational feasibility by focusing adversarial training on a lightweight projector layer instead of the entire model; additionally, we design a dynamic weight adjustment mechanism that optimizes the loss function's weight allocation based on task demands, streamlining the tuning process. To enhance defense performance, we propose a joint optimization strategy across visual and textual modalities, ensuring robust resistance to jailbreak attacks originating from either modality. Extensive experiments conducted on five major jailbreak attack methods across three mainstream MLLMs demonstrate the effectiveness of our approach. ProEAT achieves state-of-the-art defense performance, outperforming existing baselines by an average margin of +34% across text and image modalities, while incurring only a 1% reduction in clean accuracy. Furthermore, evaluations on real-world embodied intelligent systems highlight the practical applicability of our framework, paving the way for the development of more secure and reliable multimodal systems.

Graph Neural Networks (GNNs) have gained attention for their ability to learn representations from graph data. Due to privacy concerns and conflicts of interest that prevent clients from directly sharing graph data with one another, Vertical Graph Federated Learning (VGFL) frameworks have been developed. Recent studies have shown that VGFL is vulnerable to adversarial attacks that degrade performance. However, it is a common problem that client nodes are often unlabeled in the realm of VGFL. Consequently, the existing attacks, which rely on the availability of labeling information to obtain gradients, are inherently constrained in their applicability. This limitation precludes their deployment in practical, real-world environments. To address the above problems, we propose a novel graph adversarial attack against VGFL, referred to as VGFL-SA, to degrade the performance of VGFL by modifying the local clients structure without using labels. Specifically, VGFL-SA uses a contrastive learning method to complete the attack before the local clients are trained. VGFL-SA first accesses the graph structure and node feature information of the poisoned clients, and generates the contrastive views by node-degree-based edge augmentation and feature shuffling augmentation. Then, VGFL-SA uses the shared graph encoder to get the embedding of each view, and the gradients of the adjacency matrices are obtained by the contrastive function. Finally, perturbed edges are generated using gradient modification rules. We validated the performance of VGFL-SA by performing a node classification task on real-world datasets, and the results show that VGFL-SA achieves good attack effectiveness and transferability.

Out-of-Distribution (OOD) detection is critical for ensuring the reliability of machine learning models in safety-critical applications such as autonomous driving and medical diagnosis. While deploying personalized OOD detection directly on edge devices is desirable, it remains challenging due to large model sizes and the computational infeasibility of on-device training. Federated learning partially addresses this but still requires gradient computation and backpropagation, exceeding the capabilities of many edge devices. To overcome these challenges, we propose SecDOOD, a secure cloud-device collaboration framework for efficient on-device OOD detection without requiring device-side backpropagation. SecDOOD utilizes cloud resources for model training while ensuring user data privacy by retaining sensitive information on-device. Central to SecDOOD is a HyperNetwork-based personalized parameter generation module, which adapts cloud-trained models to device-specific distributions by dynamically generating local weight adjustments, effectively combining central and local information without local fine-tuning. Additionally, our dynamic feature sampling and encryption strategy selectively encrypts only the most informative feature channels, largely reducing encryption overhead without compromising detection performance. Extensive experiments across multiple datasets and OOD scenarios demonstrate that SecDOOD achieves performance comparable to fully fine-tuned models, enabling secure, efficient, and personalized OOD detection on resource-limited edge devices. To enhance accessibility and reproducibility, our code is publicly available at https://github.com/Dystopians/SecDOOD.