These same concepts were also used to label 656 skin disease images from the Diverse Dermatology Images dataset, providing an additional external dataset with diverse skin tone representations. We review the potential applications fortheSkinCon dataset, such asprobing models, concept-based explanations, concept bottlenecks, error analysis, andslice discovery.

Adversarial patch attacks pose a practical threat to deep learning models by forcing targeted misclassifications through localized perturbations, often realized in the physical world. Existing defenses typically assume prior knowledge of patch size or location, limiting their applicability. In this work, we propose a patch-agnostic defense that leverages concept-based explanations to identify and suppress the most influential concept activation vectors, thereby neutralizing patch effects without explicit detection. Evaluated on Imagenette with a ResNet-50, our method achieves higher robust and clean accuracy than the state-of-the-art PatchCleanser, while maintaining strong performance across varying patch sizes and locations. Our results highlight the promise of combining interpretability with robustness and suggest concept-driven defenses as a scalable strategy for securing machine learning models against adversarial patch attacks.

Security alignment enables the Large Language Model (LLM) to gain the protection against malicious queries, but various jailbreak attack methods reveal the vulnerability of this security mechanism. Previous studies have isolated LLM jailbreak attacks and defenses. We analyze the security protection mechanism of the LLM, and propose a framework that combines attack and defense. Our method is based on the linearly separable property of LLM intermediate layer embedding, as well as the essence of jailbreak attack, which aims to embed harmful problems and transfer them to the safe area. We utilize generative adversarial network (GAN) to learn the security judgment boundary inside the LLM to achieve efficient jailbreak attack and defense. The experimental results indicate that our method achieves an average jailbreak success rate of 88.85\% across three popular LLMs, while the defense success rate on the state-of-the-art jailbreak dataset reaches an average of 84.17\%. This not only validates the effectiveness of our approach but also sheds light on the internal security mechanisms of LLMs, offering new insights for enhancing model security The code and data are available at https://github.com/NLPGM/CAVGAN.

An essential step in deploying medical imaging models is ensuring alignment with clinical knowledge and interpretability. We focus on mapping clinical concepts into the latent space of generative models to identify Concept Activation Vectors (CAVs). Using a simple reconstruction autoencoder, we link user-defined concepts to image-level features without explicit label training. The extracted concepts are stable across datasets, enabling visual explanations that highlight clinically relevant features. By traversing latent space along concept directions, we produce counterfactuals that exaggerate or reduce specific clinical features. Preliminary results on chest X-rays show promise for large pathologies like cardiomegaly, while smaller pathologies remain challenging due to reconstruction limits. Although not outperforming baselines, this approach offers a path toward interpretable, concept-based explanations aligned with clinical knowledge.

Despite careful safety alignment, current large language models (LLMs) remain vulnerable to various attacks. To further unveil the safety risks of LLMs, we introduce a Safety Concept Activation Vector (SCAV) framework, which effectively guides the attacks by accurately interpreting LLMs' safety mechanisms. We then develop an SCAV-guided attack method that can generate both attack prompts and embedding-level attacks with automatically selected perturbation hyperparameters. Both automatic and human evaluations demonstrate that our attack method significantly improves the attack success rate and response quality while requiring less training data. Additionally, we find that our generated attack prompts may be transferable to GPT-4, and the embedding-level attacks may also be transferred to other white-box LLMs whose parameters are known.

Concept activation vector (CAV) has attracted broad research interest in explainable AI, by elegantly attributing model predictions to specific concepts. However, the training of CAV often necessitates a large number of high-quality images, which are expensive to curate and thus limited to a predefined set of concepts. To address this issue, we propose Language-Guided CAV (LG-CAV) to harness the abundant concept knowledge within the certain pre-trained vision-language models (e.g., CLIP). This method allows training any CAV without labeled data, by utilizing the corresponding concept descriptions as guidance. To bridge the gap between vision-language model and the target model, we calculate the activation values of concept descriptions on a common pool of images (probe images) with vision-language model and utilize them as language guidance to train the LG-CAV.

While advances in deep learning in the last years have led to impressive performance in many domains, such models are not always reliable, particularly when it comes to generalizing to new environments or adversarial attacks. To improve on that, numerous methods have been developed in the field of explainable artificial intelligence (xAI) [5] to provide insights into model behavior and facilitate actionable modifications. However, the majority of methods focus on explaining model predictions, which can help understand misclassifications but do not explicitly address predictive uncertainty(See Figure 1). Understanding uncertainty is crucial for detecting potential model weaknesses, particularly in dynamic environments. Since uncertainty quantification is useful in various applications, including active learning [20], classification with rejects [17], adversarial example detection [26], and reinforcement learning [24], a significant body of work aims to improve the quantification of predictive uncertainty using Bayesian deep learning (BDL) and approximations thereof [15,9,14]. In contrast, the literature on understanding the sources of uncertainty for a given model via explanations is limited, focusing on methods for feature attribution [28,27] (see section 2.4 for more related

As large language models (LLMs) are widely deployed across various domains, the ability to control their generated outputs has become more critical. This control involves aligning LLMs outputs with human values and ethical principles or customizing LLMs on specific topics or styles for individual users. Existing controlled generation methods either require significant computational resources and extensive trial-and-error or provide coarse-grained control. In this paper, we propose Generation with Concept Activation Vector (GCAV), a lightweight model control framework that ensures accurate control without requiring resource-extensive fine-tuning. Specifically, GCAV first trains a concept activation vector for specified concepts to be controlled, such as toxicity. During inference, GCAV steers the concept vector in LLMs, for example, by removing the toxicity concept vector from the activation layers. Control experiments from different perspectives, including toxicity reduction, sentiment control, linguistic style, and topic control, demonstrate that our framework achieves state-of-the-art performance with granular control, allowing for fine-grained adjustments of both the steering layers and the steering magnitudes for individual samples.

Machine learning models use high dimensional feature spaces to map their inputs to the corresponding class labels. However, these features often do not have a one-to-one correspondence with physical concepts understandable by humans, which hinders the ability to provide a meaningful explanation for the decisions made by these models. We propose a method for measuring the correlation between high-level concepts and the decisions made by a machine learning model. Our method can isolate the impact of a given high-level concept and accurately measure it quantitatively. Additionally, this study aims to determine the prevalence of frequent patterns in machine learning models, which often occur in imbalanced datasets. We have successfully applied the proposed method to fundus images and managed to quantitatively measure the impact of radiomic patterns on the model decisions.

Machine learning models are trained with relatively simple objectives, such as next token prediction. However, on deployment, they appear to capture a more fundamental representation of their input data. It is of interest to understand the nature of these representations to help interpret the model's outputs and to identify ways to improve the salience of these representations. Concept vectors are constructions aimed at attributing concepts in the input data to directions, represented by vectors, in the model's latent space. In this work, we introduce concept boundary vectors as a concept vector construction derived from the boundary between the latent representations of concepts. Empirically we demonstrate that concept boundary vectors capture a concept's semantic meaning, and we compare their effectiveness against concept activation vectors.