Unveiling Concept Attribution in Diffusion Models

• URL: http://arxiv.org/abs/2412.02542v3
• Date: Tue, 28 Oct 2025 03:07:50 GMT
• Title: Unveiling Concept Attribution in Diffusion Models
• Authors: Quang H. Nguyen, Hoang Phan, Khoa D. Doan,
• Abstract summary: Diffusion models have shown remarkable abilities in generating realistic and high-quality images from text prompts.<n>Recent works employ causal tracing to localize knowledge-storing layers in generative models without showing how other layers contribute to the target concept.<n>We decompose diffusion models using component attribution, systematically unveiling the importance of each component in generating a concept.
• Score: 12.77092262246859
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Diffusion models have shown remarkable abilities in generating realistic and high-quality images from text prompts. However, a trained model remains largely black-box; little do we know about the roles of its components in exhibiting a concept such as objects or styles. Recent works employ causal tracing to localize knowledge-storing layers in generative models without showing how other layers contribute to the target concept. In this work, we approach diffusion models' interpretability problem from a more general perspective and pose a question: \textit{``How do model components work jointly to demonstrate knowledge?''}. To answer this question, we decompose diffusion models using component attribution, systematically unveiling the importance of each component (specifically the model parameter) in generating a concept. The proposed framework, called \textbf{C}omponent \textbf{A}ttribution for \textbf{D}iffusion Model (CAD), discovers the localization of concept-inducing (positive) components, while interestingly uncovers another type of components that contribute negatively to generating a concept, which is missing in the previous knowledge localization work. Based on this holistic understanding of diffusion models, we introduce two fast, inference-time model editing algorithms, CAD-Erase and CAD-Amplify; in particular, CAD-Erase enables erasure and CAD-Amplify allows amplification of a generated concept by ablating the positive and negative components, respectively, while retaining knowledge of other concepts. Extensive experimental results validate the significance of both positive and negative components pinpointed by our framework, demonstrating the potential of providing a complete view of interpreting generative models. Our code is available \href{https://github.com/mail-research/CAD-attribution4diffusion}{here}.

Related papers

• What Drives Compositional Generalization in Visual Generative Models? [56.01574461407906]
  We conduct a systematic study of how various design choices influence compositional generalization in image and video generation.<n>We identify two key factors: (i) whether the training objective operates on a discrete or continuous distribution, and (ii) to what extent conditioning provides information about the constituent concepts during training.<n>Building on these insights, we show that relaxing the MaskGIT discrete loss with an auxiliary continuous JEPA-based objective can improve compositional performance in discrete models like MaskGIT.
  arXiv  Detail & Related papers  (2025-10-03T15:02:27Z)
• Fine-Grained Erasure in Text-to-Image Diffusion-based Foundation Models [56.35484513848296]
  FADE (Fine grained Attenuation for Diffusion Erasure) is an adjacency-aware unlearning algorithm for text-to-image generative models. It removes target concepts with minimal impact on correlated concepts, achieving a 12% improvement in retention performance over state-of-the-art methods.
  arXiv  Detail & Related papers  (2025-03-25T15:49:48Z)
• Concept Layers: Enhancing Interpretability and Intervenability via LLM Conceptualization [2.163881720692685]
  We introduce a new methodology for incorporating interpretability and intervenability into an existing model by integrating Concept Layers into its architecture. Our approach projects the model's internal vector representations into a conceptual, explainable vector space before reconstructing and feeding them back into the model. We evaluate CLs across multiple tasks, demonstrating that they maintain the original model's performance and agreement while enabling meaningful interventions.
  arXiv  Detail & Related papers  (2025-02-19T11:10:19Z)
• Scaling Concept With Text-Guided Diffusion Models [53.80799139331966]
  Instead of replacing a concept, can we enhance or suppress the concept itself? We introduce ScalingConcept, a simple yet effective method to scale decomposed concepts up or down in real input without introducing new elements. More importantly, ScalingConcept enables a variety of novel zero-shot applications across image and audio domains.
  arXiv  Detail & Related papers  (2024-10-31T17:09:55Z)
• Human-Object Interaction Detection Collaborated with Large Relation-driven Diffusion Models [65.82564074712836]
  We introduce DIFfusionHOI, a new HOI detector shedding light on text-to-image diffusion models. We first devise an inversion-based strategy to learn the expression of relation patterns between humans and objects in embedding space. These learned relation embeddings then serve as textual prompts, to steer diffusion models generate images that depict specific interactions.
  arXiv  Detail & Related papers  (2024-10-26T12:00:33Z)
• ProtoS-ViT: Visual foundation models for sparse self-explainable classifications [0.6249768559720122]
  Prototypical networks aim to build intrinsically explainable models based on the linear summation of concepts. This work first proposes an extensive set of quantitative and qualitative metrics which allow to identify drawbacks in current prototypical networks. It then introduces a novel architecture which provides compact explanations, outperforming current prototypical models in terms of explanation quality.
  arXiv  Detail & Related papers  (2024-06-14T13:36:30Z)
• Diffusion Model with Cross Attention as an Inductive Bias for Disentanglement [58.9768112704998]
  Disentangled representation learning strives to extract the intrinsic factors within observed data. We introduce a new perspective and framework, demonstrating that diffusion models with cross-attention can serve as a powerful inductive bias. This is the first work to reveal the potent disentanglement capability of diffusion models with cross-attention, requiring no complex designs.
  arXiv  Detail & Related papers  (2024-02-15T05:07:54Z)
• Advancing Ante-Hoc Explainable Models through Generative Adversarial Networks [24.45212348373868]
  This paper presents a novel concept learning framework for enhancing model interpretability and performance in visual classification tasks. Our approach appends an unsupervised explanation generator to the primary classifier network and makes use of adversarial training. This work presents a significant step towards building inherently interpretable deep vision models with task-aligned concept representations.
  arXiv  Detail & Related papers  (2024-01-09T16:16:16Z)
• Attributing Learned Concepts in Neural Networks to Training Data [5.930268338525991]
  We find evidence for convergence, where removing the 10,000 top attributing images for a concept and retraining the model does not change the location of the concept in the network. This suggests that the features that inform the development of a concept are spread in a more diffuse manner across its exemplars, implying robustness in concept formation.
  arXiv  Detail & Related papers  (2023-10-04T20:26:59Z)
• The Hidden Language of Diffusion Models [70.03691458189604]
  We present Conceptor, a novel method to interpret the internal representation of a textual concept by a diffusion model. We find surprising visual connections between concepts, that transcend their textual semantics. We additionally discover concepts that rely on mixtures of exemplars, biases, renowned artistic styles, or a simultaneous fusion of multiple meanings.
  arXiv  Detail & Related papers  (2023-06-01T17:57:08Z)
• Reduce, Reuse, Recycle: Compositional Generation with Energy-Based Diffusion Models and MCMC [102.64648158034568]
  diffusion models have quickly become the prevailing approach to generative modeling in many domains. We propose an energy-based parameterization of diffusion models which enables the use of new compositional operators. We find these samplers lead to notable improvements in compositional generation across a wide set of problems.
  arXiv  Detail & Related papers  (2023-02-22T18:48:46Z)
• Compositional Visual Generation with Composable Diffusion Models [80.75258849913574]
  We propose an alternative structured approach for compositional generation using diffusion models. An image is generated by composing a set of diffusion models, with each of them modeling a certain component of the image. The proposed method can generate scenes at test time that are substantially more complex than those seen in training.
  arXiv  Detail & Related papers  (2022-06-03T17:47:04Z)
• Model LEGO: Creating Models Like Disassembling and Assembling Building Blocks [53.09649785009528]
  In this paper, we explore a paradigm that does not require training to obtain new models. Similar to the birth of CNN inspired by receptive fields in the biological visual system, we propose Model Disassembling and Assembling. For model assembling, we present the alignment padding strategy and parameter scaling strategy to construct a new model tailored for a specific task.
  arXiv  Detail & Related papers  (2022-03-25T05:27:28Z)
• Translational Concept Embedding for Generalized Compositional Zero-shot Learning [73.60639796305415]
  Generalized compositional zero-shot learning means to learn composed concepts of attribute-object pairs in a zero-shot fashion. This paper introduces a new approach, termed translational concept embedding, to solve these two difficulties in a unified framework.
  arXiv  Detail & Related papers  (2021-12-20T21:27:51Z)
• Relation-aware Compositional Zero-shot Learning for Attribute-Object Pair Recognition [17.464548471883948]
  This paper proposes a novel model for recognizing images with composite attribute-object concepts. We aim to explore the three key properties required to learn rich and robust features for primitive concepts that compose attribute-object pairs. To prevent the model from being biased towards seen composite concepts and reduce the entanglement between attributes and objects, we propose a blocking mechanism.
  arXiv  Detail & Related papers  (2021-08-10T11:23:03Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.