Concept erasure has emerged as a promising technique for mitigating the risk of harmful content generation in diffusion models by selectively unlearning undesirable concepts. The common principle of previous works to remove a specific concept is to map it to a fixed generic concept, such as a neutral concept or just an empty text prompt. In this paper, we demonstrate that this fixed-target strategy is suboptimal, as it fails to account for the impact of erasing one concept on the others. To address this limitation, we model the concept space as a graph and empirically analyze the effects of erasing one concept on the remaining concepts. Our analysis uncovers intriguing geometric properties of the concept space, where the influence of erasing a concept is confined to a local region. Building on this insight, we propose the Adaptive Guided Erasure (AGE) method, which \emph{dynamically} selects optimal target concepts tailored to each undesirable concept, minimizing unintended side effects. Experimental results show that AGE significantly outperforms state-of-the-art erasure methods on preserving unrelated concepts while maintaining effective erasure performance. Our code is published at {https://github.com/tuananhbui89/Adaptive-Guided-Erasure}.

The personalization techniques of diffusion models succeed in generating specific concepts but also pose threats to copyright protection and illegal use. Model Watermarking is an effective method to prevent the unauthorized use of subject-driven or style-driven image generation, safeguarding concept copyrights. However, under the goal of concept-oriented protection, current watermarking schemes typically add watermarks to all images rather than applying them in a refined manner targeted at specific concepts. Additionally, the personalization techniques of diffusion models can easily remove watermarks. Existing watermarking methods struggle to achieve fine-grained watermark embedding with a few images of specific concept and prevent removal of watermarks through personalized fine-tuning. Therefore, we introduce a novel concept-oriented watermarking framework that seamlessly embeds imperceptible watermarks into the concept of diffusion models. We conduct extensive experiments and ablation studies to verify our framework. Our code is available at https://anonymous.4open.science/r/Conceptwm-4EB3/.

With the advancement of computer vision and natural language processing, text-to-video generation, enabled by text-to-video diffusion models, has become more prevalent. These models are trained using a large amount of data from the internet. However, the training data often contain copyrighted content, including cartoon character icons and artist styles, private portraits, and unsafe videos. Since filtering the data and retraining the model is challenging, methods for unlearning specific concepts from text-to-video diffusion models have been investigated. However, due to the high computational complexity and relative large optimization scale, there is little work on unlearning methods for text-to-video diffusion models. We propose a novel concept-unlearning method by transferring the unlearning capability of the text encoder of text-to-image diffusion models to text-to-video diffusion models. Specifically, the method optimizes the text encoder using few-shot unlearning, where several generated images are used. We then use the optimized text encoder in text-to-video diffusion models to generate videos. Our method costs low computation resources and has small optimization scale. We discuss the generated videos after unlearning a concept. The experiments demonstrates that our method can unlearn copyrighted cartoon characters, artist styles, objects and people's facial characteristics. Our method can unlearn a concept within about 100 seconds on an RTX 3070. Since there was no concept unlearning method for text-to-video diffusion models before, we make concept unlearning feasible and more accessible in the text-to-video domain.

Today's artificial intelligence is often described as a "black box." AI developers don't write explicit rules for these systems; instead, they feed in vast quantities of data and the systems learn on their own to spot patterns. But the inner workings of the AI models remain opaque, and efforts to peer inside them to check exactly what is happening haven't progressed very far. Beneath the surface, neural networks--today's most powerful type of AI--consist of billions of artificial "neurons" represented as decimal-point numbers. Nobody truly understands what they mean, or how they work.

For the past decade, AI researcher Chris Olah has been obsessed with artificial neural networks. One question in particular engaged him, and has been the center of his work, first at Google Brain, then OpenAI, and today at AI startup Anthropic, where he is a cofounder. "What's going on inside of them?" he says. "We have these systems, we don't know what's going on. That question has become a core concern now that generative AI has become ubiquitous. Large language models like ChatGPT, Gemini, and Anthropic's own Claude have dazzled people with their language prowess and infuriated people with their tendency to make things up. Their potential to solve previously intractable problems enchants techno-optimists. But LLMs are strangers in our midst. Even the people who build them don't know exactly how they work, and massive effort is required to create guardrails to prevent them from churning out bias, misinformation, and even blueprints for deadly chemical weapons. If the people building the models knew what happened inside these "black boxes,'' it would be easier to make them safer.

In this work we study the presence of expert units in pre-trained Transformer Models (TM), and how they impact a model's performance. We define expert units to be neurons that are able to classify a concept with a given average precision, where a concept is represented by a binary set of sentences containing the concept (or not). Leveraging the OneSec dataset (Scarlini et al., 2019), we compile a dataset of 1641 concepts that allows diverse expert units in TM to be discovered. We show that expert units are important in several ways: (1) The presence of expert units is correlated ($r^2=0.833$) with the generalization power of TM, which allows ranking TM without requiring fine-tuning on suites of downstream tasks. We further propose an empirical method to decide how accurate such experts should be to evaluate generalization. (2) The overlap of top experts between concepts provides a sensible way to quantify concept co-learning, which can be used for explainability of unknown concepts. (3) We show how to self-condition off-the-shelf pre-trained language models to generate text with a given concept by forcing the top experts to be active, without requiring re-training the model or using additional parameters.

Efficiently querying Description Logic (DL) ontologies is becoming a vital task in various data-intensive DL applications. Considered as a basic service for answering object queries over DL ontologies, instance checking can be realized by using the most specific concept (MSC) method, which converts instance checking into subsumption problems. This method, however, loses its simplicity and efficiency when applied to large and complex ontologies, as it tends to generate very large MSC's that could lead to intractable reasoning. In this paper, we propose a revision to this MSC method for DL SHI, allowing it to generate much simpler and smaller concepts that are specific-enough to answer a given query. With independence between computed MSC's, scalability for query answering can also be achieved by distributing and parallelizing the computations. An empirical evaluation shows the efficacy of our revised MSC method and the significant efficiency achieved when using it for answering object queries.

Instance checking is considered a central service for data retrieval from description logic (DL) ontologies. In this paper, we propose a revised most specific concept (MSC) method for DL SHI}, which converts instance checking into subsumption problems. This revised method can generate small concepts that are specific-enough to answer a given query, and allow reasoning to explore only a subset of the ABox data to achieve efficiency. Experiments show effectiveness of our proposed method in terms of concept size reduction and the improvement in reasoning efficiency.