The success of diffusion models has raised concerns about the generation of unsafe or harmful content, prompting concept erasure approaches that fine-tune modules to suppress specific concepts while preserving general generative capabilities. However, as the number of erased concepts grows, these methods often become inefficient and ineffective, since each concept requires a separate set of fine-tuned parameters and may degrade the overall generation quality. In this work, we propose a supertype-subtype concept hierarchy that organizes erased concepts into a parent–child structure. Each erased concept is treated as a child node, and semantically related concepts (e.g., macaw, and bald eagle) are grouped under a shared parent node, referred to as a supertype concept (e.g., bird). Rather than erasing concepts individually, we introduce an effective and efficient group-wise suppression method, where semantically similar concepts are grouped and erased jointly by sharing a single set of learnable parameters. During the erasure phase, standard diffusion regularization is applied to preserve denoising process in unmasked regions. To mitigate the degradation of supertype generation caused by excessive erasure of semantically related subtypes, we propose a novel method called Supertype-Preserving Low-Rank Adaptation (SuPLoRA), which encodes the supertype concept information in the frozen down-projection matrix and updates only the up-projection matrix during erasure. Theoretical analysis demonstrates the effectiveness of SuPLoRA in mitigating generation performance degradation. We construct a more challenging benchmark that requires simultaneous erasure of concepts across diverse domains, including celebrities, objects, and pornographic content. Comprehensive experiments demonstrate that our method achieves a superior balance between effective multi-concept erasure and the preservation of desirable generative performance.

Tactile sensation plays a crucial role in the development of multi-modal large models and embodied intelligence. To collect tactile data with minimal cost as possible, a series of studies have attempted to generate tactile images by vision-to-touch image translation. However, compared to text modality, visual modality-driven tactile generation cannot accurately depict human tactile sensation. In this work, we analyze the characteristics of tactile images in detail from two granularities: object-level (tactile texture, tactile shape), and sensor-level (gel status). We model these granularities of information through text descriptions and propose a fine-grained Text-to-Touch generation method (TextToucher) to generate high-quality tactile samples. Specifically, we introduce a multimodal large language model to build the text sentences about object-level tactile information and employ a set of learnable text prompts to represent the sensor-level tactile information. To better guide the tactile generation process with the built text information, we fuse the dual grains of text information and explore various dual-grain text conditioning methods within the diffusion transformer architecture. Furthermore, we propose a Contrastive Text-Touch Pre-training (CTTP) metric to precisely evaluate the quality of text-driven generated tactile data. Extensive experiments demonstrate the superiority of our TextToucher method.

Learning from data streams collected sequentially over time are widely spread in real-world applications. Previous methods typically assume that the data stream has a feature space with a fixed or clearly defined evolution pattern, as well as a balanced class distribution. However, in many practical scenarios, such as environmental monitoring systems, the frequency of anomalous events is significantly imbalanced compared to normal ones and the feature space dynamically changes due to ecological evolution and sensor lifespan. To alleviate this important but rarely studied problem, we propose the Adaptive Learning in Imbalace data streams with Unpredictable feature evolution (ALIU) algorithm. As data streams with imbalanced class distribution arrive, ALIU first mitigates the model's bias for the majority class by reweighting the adaptive gradient descent magnitudes between different classes. Then, a new loss function is proposed that simultaneously focuses on misclassifications and maintains model robustness. Further, when imbalanced data streams arrive with feature evolutions, we reuse the previously learned model and update the incomplete and augmented features by adopting the adaptive gradient strategy and ensemble method, respectively. Finally, we utilize the projected technique to build a sparse yet efficient model. Based on a few common and mild assumptions, we theoretically analyze that the ALIU satisfies a sub-linear regret bound under both convex and strong convex loss functions and the performance of model can be improved with the assistance of old features. Besides, extensive experimental results further demonstrate the effectiveness of our proposed algorithm.

In autonomous driving, deep models have shown remarkable performance across various visual perception tasks with the demand of high-quality and huge-diversity training datasets. Such datasets are expected to cover various driving scenarios with adverse weather, lighting conditions, and diverse moving objects. However, manually collecting these data presents huge challenges and is expensive. With the rapid development of large generative models, we propose DriveDiTFit, a novel method for efficiently generating autonomous Driv ing data by Fi ne- t uning pre-trained Di ffusion T ransformers (DiTs). Specifically, DriveDiTFit utilizes a gap-driven modulation technique to carefully select and efficiently fine-tune a few parameters in DiTs according to the discrepancy between the pre-trained source data and the target driving data. Additionally, DriveDiTFit develops an effective weather and lighting condition embedding module to ensure diversity in the generated data, which is initialized by a nearest-semantic-similarity initialization approach. Through progressive tuning scheme to refine the process of detail generation in early diffusion process and enlarging the weights corresponding to small objects in training loss, DriveDiTFit ensures high-quality generation of small moving objects in the generated data. Extensive experiments conducted on driving datasets confirm that our method could efficiently produce diverse real driving data.

Social phobia is a prevalent mental health con-dition characterized by overwhelming fear on apprehension of social situations, often leading individuals to avoid such encounters. It is a very common mental disorder among adolescents and can have a significant bad effect on their social skills and well-beings if not intervened early. Existing methods to assess social anxiety level rely on self-report scales or clinical diagnosis, both of which may face challenges like subjective judgments and patient resistance. To address this, we applied Virtual Reality technology (VR) and introduced event-oriented social anxiety dataset (EOSAD) to serve for fine-grained social anxiety level detected model. A case study (N = 44) was conducted where participants wore a Vive Pro Eye HMD to experience two types of scenes and labeled their social anxiety level accordingly. During the experiment, relevant signals and social anxiety scores were collected, including (1) behavioral signals (head and eye movements) (2) physiological signals (heart rate (HR), electrodermal activity (EDA), etc.) (3) discrete social anxiety self-reported labels (4) evaluation questionnaires. We first verified the labels, and furthermore ran baseline classification experiments, where GRU model showed best accuracy: 92.84 %for binary classification. It was also found that either behavioral data or physiological signals alone could achieve satisfactory accuracies (85.64% and 91.02%) while the combined achieved slightly higher(92.84 %)