# Why Do Video Diffusion Models Violate Physics? Unveiling the Flaws in Attention Mechanisms

arXiv: https://arxiv.org/abs/2609.23658 · HTML: https://mlsanity.com/paper/2609.23658/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in Video generation, one of 14 directions in Diffusion models.
- Builds on 38 of its 43 references that are in the map.
- First posted on arXiv 20 Sep 2026.

## Placed in

- Diffusion models / [Video generation](https://mlsanity.com/diffusion/video-generation/)

## Builds on (most cited first, 30 of 38)

- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Wan: Open and Advanced Large-Scale Video Generative Models](https://mlsanity.com/paper/2503.20314/) (2025)
- [Transformer Feed-Forward Layers Are Key-Value Memories](https://mlsanity.com/paper/2012.14913/) (2020)
- [Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small](https://mlsanity.com/paper/2211.00593/) (2022)
- [WorldSimBench: Towards Video Generation Models as World Simulators](https://mlsanity.com/paper/2410.18072/) (2024)
- [VideoMAE V2: Scaling Video Masked Autoencoders with Dual Masking](https://mlsanity.com/paper/2303.16727/) (2023)
- [What the DAAM: Interpreting Stable Diffusion Using Cross Attention](https://mlsanity.com/paper/2210.04885/) (2022)
- [VideoPhy: Evaluating Physical Commonsense for Video Generation](https://mlsanity.com/paper/2406.03520/) (2024)
- [How Far is Video Generation from World Model: A Physical Law Perspective](https://mlsanity.com/paper/2411.02385/) (2024)
- [PhysGen: Rigid-Body Physics-Grounded Image-to-Video Generation](https://mlsanity.com/paper/2409.18964/) (2024)
- [Towards World Simulator: Crafting Physical Commonsense-Based Benchmark for Video Generation](https://mlsanity.com/paper/2410.05363/) (2024)
- [Seedance 2.0: Advancing Video Generation for World Complexity](https://mlsanity.com/paper/2604.14148/) (2026)
- [Stable Flow: Vital Layers for Training-Free Image Editing](https://mlsanity.com/paper/2411.14430/) (2024)
- [VideoREPA: Learning Physics for Video Generation through Relational Alignment with Foundation Models](https://mlsanity.com/paper/2505.23656/) (2025)
- [WISA: World Simulator Assistant for Physics-Aware Text-to-Video Generation](https://mlsanity.com/paper/2503.08153/) (2025)
- [PhyT2V: LLM-Guided Iterative Self-Refinement for Physics-Grounded Text-to-Video Generation](https://mlsanity.com/paper/2412.00596/) (2024)
- [SAeUron: Interpretable Concept Unlearning in Diffusion Models with Sparse Autoencoders](https://mlsanity.com/paper/2501.18052/) (2025)
- [GPT4Motion: Scripting Physical Motions in Text-to-Video Generation via Blender-Oriented GPT Planning](https://mlsanity.com/paper/2311.12631/) (2023)
- [Localizing and Editing Knowledge in Text-to-Image Generative Models](https://mlsanity.com/paper/2310.13730/) (2023)
- [VLIPP: Towards Physically Plausible Video Generation with Vision and Language Informed Physical Prior](https://mlsanity.com/paper/2503.23368/) (2025)
- [MotionCraft: Physics-based Zero-Shot Video Generation](https://mlsanity.com/paper/2405.13557/) (2024)
- [Towards Understanding the Working Mechanism of Text-to-Image Diffusion Model](https://mlsanity.com/paper/2405.15330/) (2024)
- [NewtonGen: Physics-Consistent and Controllable Text-to-Video Generation via Neural Newtonian Dynamics](https://mlsanity.com/paper/2509.21309/) (2025)
- [Inference-time Physics Alignment of Video Generative Models with Latent World Models](https://mlsanity.com/paper/2601.10553/) (2026)
- [Emergent Temporal Correspondences from Video Diffusion Transformers](https://mlsanity.com/paper/2506.17220/) (2025)
- [Generative Physical AI in Vision: A Survey](https://mlsanity.com/paper/2501.10928/) (2025)
- [Emergence and Evolution of Interpretable Concepts in Diffusion Models](https://mlsanity.com/paper/2504.15473/) (2025)
- [TIDE : Temporal-Aware Sparse Autoencoders for Interpretable Diffusion Transformers in Image Generation](https://mlsanity.com/paper/2503.07050/) (2025)

## Built on by (most cited first, 2 of 2)

- [Does Physics Live in the Activations? Localizing Physical Quantities in Video Diffusion Models](https://mlsanity.com/paper/2610.03154/) (2026)
- [How Does Geometry Enter Generated Motion?](https://mlsanity.com/paper/2610.05135/) (2026)

## Abstract (arXiv)

Despite impressive visual quality, state-of-the-art video diffusion models often generate content that violates real-world physical laws. While existing solutions rely on external priors or specialized data, we investigate the root cause by exploring the internal mechanisms of these models. Specifically, we present the first interpretability study on the ''motion planning'' process of text-to-video diffusion models, revealing how motion trajectories form during early denoising stages. Building upon the ''first shape, then details'' finding, we combine cross-attention trajectory patterns with causal head contributions to identify a specific subset of attention heads driving motion planning. Further, our self-attention analysis shows that Rotary Position Embedding (RoPE) induces excessive spatial attention decay. This causes early candidate regions to prematurely lock into physically implausible positions, suppressing reasonable trajectories in adjacent frames and triggering generation failure modes. To address this fundamental flaw, we propose a lightweight architectural modification that scales the frequency of RoPE across different denoising steps. This strategy reduces excessive attention decay, helping the model explore better candidate regions to establish coherent physical motion. Finally, training-free and training-based experiments confirm the effectiveness of our approach in enhancing the physical commonsense of generated videos.