AttentionBender: Manipulating Cross-Attention in Video Diffusion Transformers as a Creative Probe

• URL: http://arxiv.org/abs/2604.20936v1
• Date: Wed, 22 Apr 2026 13:11:21 GMT
• Title: AttentionBender: Manipulating Cross-Attention in Video Diffusion Transformers as a Creative Probe
• Abstract summary: AttentionBender is a tool that manipulates cross-attention in Video Diffusion Transformers.<n>We assess AttentionBender by visualizing 4,500+ video generations across prompts, operations, and layer targets.
• Score: 0.03437656066916039
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We present AttentionBender, a tool that manipulates cross-attention in Video Diffusion Transformers to help artists probe the internal mechanics of black-box video generation. While generative outputs are increasingly realistic, prompt-only control limits artists' ability to build intuition for the model's material process or to work beyond its default tendencies. Using an autobiographical research-through-design approach, we built on Network Bending to design AttentionBender, which applies 2D transforms (rotation, scaling, translation, etc.) to cross-attention maps to modulate generation. We assess AttentionBender by visualizing 4,500+ video generations across prompts, operations, and layer targets. Our results suggest that cross-attention is highly entangled: targeted manipulations often resist clean, localized control, producing distributed distortions and glitch aesthetics over linear edits. AttentionBender contributes a tool that functions both as an Explainable AI style probe of transformer attention mechanisms, and as a creative technique for producing novel aesthetics beyond the model's learned representational space.

Related papers

• Real-Time AttentionBender: Granular Interactive Network Bending of Video Diffusion Transformers [2.0697105762666324]
  Real-Time AttentionBender is a tool that extends the practice of network bending across the full depth of the video diffusion transformer (DiT)<n>It exposes self-attention, cross-attention, and the feed-forward network as independently manipulable surfaces.<n>The immediacy of live manipulation affords what we call "material intimacy" with the model.
  arXiv  Detail & Related papers  (2026-04-24T12:40:58Z)
• MagicQuillV2: Precise and Interactive Image Editing with Layered Visual Cues [106.02577891104079]
  We propose MagicQuill V2, a novel system that introduces a textbflayered composition paradigm to generative image editing.<n>Our method deconstructs creative intent into a stack of controllable visual cues.
  arXiv  Detail & Related papers  (2025-12-02T18:59:58Z)
• Bitrate-Controlled Diffusion for Disentangling Motion and Content in Video [38.71994714429696]
  We propose a novel and general framework to disentangle video data into its dynamic motion and static content components.<n>Our proposed method is a self-supervised pipeline with less assumptions and inductive biases than previous works.<n>We validate our disentangled representation learning framework on real-world talking head videos with motion transfer and auto-regressive motion generation tasks.
  arXiv  Detail & Related papers  (2025-09-10T08:14:45Z)
• Attention of a Kiss: Exploring Attention Maps in Video Diffusion for XAIxArts [0.03437656066916039]
  This study proposes a method for extracting and visualizing cross-attention maps in generative video models.<n>Our tool provides an interpretable window into the temporal and spatial behavior of attention in text-to-video generation.
  arXiv  Detail & Related papers  (2025-08-30T19:46:18Z)
• Generative AI for Cel-Animation: A Survey [59.20171452237911]
  GenAI is revolutionizing traditional animation by lowering technical barriers, broadening accessibility for a wider range of creators.<n>Despite its potential, challenges like consistency, stylistic coherence, and ethical considerations persist.<n>This paper explores future directions advancements in AI-assisted animation.
  arXiv  Detail & Related papers  (2025-01-08T20:57:39Z)
• Generative Rendering: Controllable 4D-Guided Video Generation with 2D Diffusion Models [40.71940056121056]
  We present a novel approach that combines the controllability of dynamic 3D meshes with the expressivity and editability of emerging diffusion models. We demonstrate our approach on various examples where motion can be obtained by animating rigged assets or changing the camera path.
  arXiv  Detail & Related papers  (2023-12-03T14:17:11Z)
• Vision Transformers Need Registers [26.63912173005165]
  We identify and characterize artifacts in feature maps of both supervised and self-supervised ViT networks. We show that this solution fixes that problem entirely for both supervised and self-supervised models.
  arXiv  Detail & Related papers  (2023-09-28T16:45:46Z)
• AttentionViz: A Global View of Transformer Attention [60.82904477362676]
  We present a new visualization technique designed to help researchers understand the self-attention mechanism in transformers. The main idea behind our method is to visualize a joint embedding of the query and key vectors used by transformer models to compute attention. We create an interactive visualization tool, AttentionViz, based on these joint query-key embeddings.
  arXiv  Detail & Related papers  (2023-05-04T23:46:49Z)
• Holistically Explainable Vision Transformers [136.27303006772294]
  We propose B-cos transformers, which inherently provide holistic explanations for their decisions. Specifically, we formulate each model component - such as the multi-layer perceptrons, attention layers, and the tokenisation module - to be dynamic linear. We apply our proposed design to Vision Transformers (ViTs) and show that the resulting models, dubbed Bcos-ViTs, are highly interpretable and perform competitively to baseline ViTs.
  arXiv  Detail & Related papers  (2023-01-20T16:45:34Z)
• Blending Anti-Aliasing into Vision Transformer [57.88274087198552]
  discontinuous patch-wise tokenization process implicitly introduces jagged artifacts into attention maps. Aliasing effect occurs when discrete patterns are used to produce high frequency or continuous information, resulting in the indistinguishable distortions. We propose a plug-and-play Aliasing-Reduction Module(ARM) to alleviate the aforementioned issue.
  arXiv  Detail & Related papers  (2021-10-28T14:30:02Z)
• Generic Attention-model Explainability for Interpreting Bi-Modal and Encoder-Decoder Transformers [78.26411729589526]
  We propose the first method to explain prediction by any Transformer-based architecture. Our method is superior to all existing methods which are adapted from single modality explainability.
  arXiv  Detail & Related papers  (2021-03-29T15:03:11Z)

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.