# Exposing and Defending the Achilles' Heel of Video Mixture-of-Experts

arXiv: https://arxiv.org/abs/2602.01369 · HTML: https://mlsanity.com/paper/2602.01369/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Vision mixture-of-experts, one of 12 directions in Mixture-of-experts.
- Builds on 46 of its 65 references that are in the map.
- First posted on arXiv 1 Feb 2026.

## Placed in

- Mixture-of-experts / [Vision mixture-of-experts](https://mlsanity.com/moe/vision-mixture-of-experts/)

## Builds on (most cited first, 30 of 46)

- [Explaining and Harnessing Adversarial Examples](https://mlsanity.com/paper/1412.6572/) (2014)
- [Towards Deep Learning Models Resistant to Adversarial Attacks](https://mlsanity.com/paper/1706.06083/) (2017)
- [Outrageously Large Neural Networks: The Sparsely-Gated Mixture-of-Experts Layer](https://mlsanity.com/paper/1701.06538/) (2017)
- [Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity](https://mlsanity.com/paper/2101.03961/) (2021)
- [SlowFast Networks for Video Recognition](https://mlsanity.com/paper/1812.03982/) (2018)
- [Reliable evaluation of adversarial robustness with an ensemble of diverse parameter-free attacks](https://mlsanity.com/paper/2003.01690/) (2020)
- [TSM: Temporal Shift Module for Efficient Video Understanding](https://mlsanity.com/paper/1811.08383/) (2018)
- [Ensemble Adversarial Training: Attacks and Defenses](https://mlsanity.com/paper/1705.07204/) (2017)
- [Fast is better than free: Revisiting adversarial training](https://mlsanity.com/paper/2001.03994/) (2020)
- [Can Spatiotemporal 3D CNNs Retrace the History of 2D CNNs and ImageNet?](https://mlsanity.com/paper/1711.09577/) (2017)
- [Mitigating Adversarial Effects Through Randomization](https://mlsanity.com/paper/1711.01991/) (2017)
- [Training robust neural networks using Lipschitz bounds](https://mlsanity.com/paper/2005.02929/) (2020)
- [Sparse Adversarial Perturbations for Videos](https://mlsanity.com/paper/1803.02536/) (2018)
- [VideoLLM-MoD: Efficient Video-Language Streaming with Mixture-of-Depths Vision Computation](https://mlsanity.com/paper/2408.16730/) (2024)
- [Multi-Modal Data-Efficient 3D Scene Understanding for Autonomous Driving](https://mlsanity.com/paper/2405.05258/) (2024)
- [Rethinking Lipschitz Neural Networks and Certified Robustness: A Boolean Function Perspective](https://mlsanity.com/paper/2210.01787/) (2022)
- [Heuristic Black-box Adversarial Attacks on Video Recognition Models](https://mlsanity.com/paper/1911.09449/) (2019)
- [Stochastic Self-Guidance for Training-Free Enhancement of Diffusion Models](https://mlsanity.com/paper/2508.12880/) (2025)
- [Robust Mixture-of-Expert Training for Convolutional Neural Networks](https://mlsanity.com/paper/2308.10110/) (2023)
- [Isometric 3D Adversarial Examples in the Physical World](https://mlsanity.com/paper/2210.15291/) (2022)
- [Mixture of Nested Experts: Adaptive Processing of Visual Tokens](https://mlsanity.com/paper/2407.19985/) (2024)
- [DAST: Context-Aware Compression in LLMs via Dynamic Allocation of Soft Tokens](https://mlsanity.com/paper/2502.11493/) (2025)
- [Boosting the Transferability of Video Adversarial Examples via Temporal Translation](https://mlsanity.com/paper/2110.09075/) (2021)
- [Efficient Robustness Assessment via Adversarial Spatial-Temporal Focus on Videos](https://mlsanity.com/paper/2301.00896/) (2023)
- [Taming Preference Mode Collapse via Directional Decoupling Alignment in Diffusion Reinforcement Learning](https://mlsanity.com/paper/2512.24146/) (2025)
- [Temporal Coherent Test-Time Optimization for Robust Video Classification](https://mlsanity.com/paper/2302.14309/) (2023)
- [Boosting Black-box Attack to Deep Neural Networks with Conditional Diffusion Models](https://mlsanity.com/paper/2310.07492/) (2023)
- [ZeroTuning: Unlocking the Initial Token's Power to Enhance Large Language Models Without Training](https://mlsanity.com/paper/2505.11739/) (2025)
- [Analysis and Extensions of Adversarial Training for Video Classification](https://mlsanity.com/paper/2206.07953/) (2022)
- [Efficient Black-box Adversarial Attacks via Bayesian Optimization Guided by a Function Prior](https://mlsanity.com/paper/2405.19098/) (2024)

## Built on by (most cited first, 3 of 3)

- [LUVE : Latent-Cascaded Ultra-High-Resolution Video Generation with Dual Frequency Experts](https://mlsanity.com/paper/2602.11564/) (2026)
- [From Zero to Detail: A Progressive Spectral Decoupling Paradigm for UHD Image Restoration with New Benchmark](https://mlsanity.com/paper/2604.15654/) (2026)
- [Hierarchical Anti-Aesthetics: Protecting Facial Privacy against Customized Diffusion Models](https://mlsanity.com/paper/2607.02038/) (2026)

## Abstract (arXiv)

Mixture-of-Experts (MoE) has demonstrated strong performance in video understanding tasks, yet its adversarial robustness remains underexplored. Existing attack methods often treat MoE as a unified architecture, overlooking the independent and collaborative weaknesses of key components such as routers and expert modules. To fill this gap, we propose Temporal Lipschitz-Guided Attacks (TLGA) to thoroughly investigate component-level vulnerabilities in video MoE models. We first design attacks on the router, revealing its independent weaknesses. Building on this, we introduce Joint Temporal Lipschitz-Guided Attacks (J-TLGA), which collaboratively perturb both routers and experts. This joint attack significantly amplifies adversarial effects and exposes the Achilles' Heel (collaborative weaknesses) of the MoE architecture. Based on these insights, we further propose Joint Temporal Lipschitz Adversarial Training (J-TLAT). J-TLAT performs joint training to further defend against collaborative weaknesses, enhancing component-wise robustness. Our framework is plug-and-play and reduces inference cost by more than 60% compared with dense models. It consistently enhances adversarial robustness across diverse datasets and architectures, effectively mitigating both the independent and collaborative weaknesses of MoE.