# VideoMLA: Low-Rank Latent KV Cache for Minute-Scale Autoregressive Video Diffusion

arXiv: https://arxiv.org/abs/2605.30351 · HTML: https://mlsanity.com/paper/2605.30351/

## Key facts

- Cited by 6 papers in the map (data through 14 Oct 2026).
- Placed in Video generation, one of 14 directions in Diffusion models.
- Builds on 29 of its 31 references that are in the map.
- First posted on arXiv 28 May 2026.

## Placed in

- Diffusion models / [Video generation](https://mlsanity.com/diffusion/video-generation/)
- Efficient LLM inference and serving / [Fast autoregressive visual generation](https://mlsanity.com/inference/fast-autoregressive-visual-generation/)

## Builds on (most cited first, 29 of 29)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [Wan: Open and Advanced Large-Scale Video Generative Models](https://mlsanity.com/paper/2503.20314/) (2025)
- [GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints](https://mlsanity.com/paper/2305.13245/) (2023)
- [DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model](https://mlsanity.com/paper/2405.04434/) (2024)
- [Improved Distribution Matching Distillation for Fast Image Synthesis](https://mlsanity.com/paper/2405.14867/) (2024)
- [Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion](https://mlsanity.com/paper/2506.08009/) (2025)
- [From Slow Bidirectional to Fast Autoregressive Video Diffusion Models](https://mlsanity.com/paper/2412.07772/) (2024)
- [Pyramidal Flow Matching for Efficient Video Generative Modeling](https://mlsanity.com/paper/2410.05954/) (2024)
- [OpenVid-1M: A Large-Scale High-Quality Dataset for Text-to-video Generation](https://mlsanity.com/paper/2407.02371/) (2024)
- [LongLive: Real-time Interactive Long Video Generation](https://mlsanity.com/paper/2509.22622/) (2025)
- [StreamingT2V: Consistent, Dynamic, and Extendable Long Video Generation from Text](https://mlsanity.com/paper/2403.14773/) (2024)
- [Rolling Forcing: Autoregressive Long Video Diffusion in Real Time](https://mlsanity.com/paper/2509.25161/) (2025)
- [Self-Forcing++: Towards Minute-Scale High-Quality Video Generation](https://mlsanity.com/paper/2510.02283/) (2025)
- [Autoregressive Video Generation without Vector Quantization](https://mlsanity.com/paper/2412.14169/) (2024)
- [Causal Forcing: Autoregressive Diffusion Distillation Done Right for High-Quality Real-Time Interactive Video Generation](https://mlsanity.com/paper/2602.02214/) (2026)
- [SANA-Video: Efficient Video Generation with Block Linear Diffusion Transformer](https://mlsanity.com/paper/2509.24695/) (2025)
- [Infinity-RoPE: Action-Controllable Infinite Video Generation Emerges From Autoregressive Self-Rollout](https://mlsanity.com/paper/2511.20649/) (2025)
- [Reward Forcing: Efficient Streaming Video Generation with Rewarded Distribution Matching Distillation](https://mlsanity.com/paper/2512.04678/) (2025)
- [Deep Forcing: Training-Free Long Video Generation with Deep Sink and Participative Compression](https://mlsanity.com/paper/2512.05081/) (2025)
- [VideoSSM: Autoregressive Long Video Generation with Hybrid State-Space Memory](https://mlsanity.com/paper/2512.04519/) (2025)
- [Rolling Sink: Bridging Limited-Horizon Training and Open-Ended Testing in Autoregressive Video Diffusion](https://mlsanity.com/paper/2602.07775/) (2026)
- [Towards Economical Inference: Enabling DeepSeek's Multi-Head Latent Attention in Any Transformer-based LLMs](https://mlsanity.com/paper/2502.14837/) (2025)
- [MemRoPE: Training-Free Infinite Video Generation via Evolving Memory Tokens](https://mlsanity.com/paper/2603.12513/) (2026)
- [Relax Forcing: Relaxed KV-Memory for Consistent Long Video Generation](https://mlsanity.com/paper/2603.21366/) (2026)
- [TransMLA: Multi-Head Latent Attention Is All You Need](https://mlsanity.com/paper/2502.07864/) (2025)
- [LongVie: Multimodal-Guided Controllable Ultra-Long Video Generation](https://mlsanity.com/paper/2508.03694/) (2025)
- [Anchor Forcing: Anchor Memory and Tri-Region RoPE for Interactive Streaming Video Diffusion](https://mlsanity.com/paper/2603.13405/) (2026)
- [Causality in Video Diffusers is Separable from Denoising](https://mlsanity.com/paper/2602.10095/) (2026)
- [Multi-head Temporal Latent Attention](https://mlsanity.com/paper/2505.13544/) (2025)

## Built on by (most cited first, 6 of 6)

- [DensityKV: Density-Guided KV Cache Compression for Long Video Generation](https://mlsanity.com/paper/2608.27922/) (2026)
- [DeCoPrune: Efficient KV-Cache Pruning for Autoregressive Video Diffusion via Denoising Consistency](https://mlsanity.com/paper/2609.39096/) (2026)
- [HLA-WM: Hybrid Linear Attention for Long-Horizon Video World Models](https://mlsanity.com/paper/2610.05739/) (2026)
- [Freeze the Decoder, Heal the Encoder: Parameter-Efficient Adaptation for SVD-Based KV-Cache Compression](https://mlsanity.com/paper/2610.10552/) (2026)
- [CORAL-KV: Compact Cross-Layer Prediction and Residual Coding for Autoregressive Video KV Caches](https://mlsanity.com/paper/2610.12514/) (2026)
- [MLAfication: Efficiently Enabling High-Fidelity MLA for Modern LLMs](https://mlsanity.com/paper/2610.17202/) (2026)

## Abstract (arXiv)

Long-rollout causal video diffusion has converged on a fixed-size sliding-window KV cache, with recent progress innovating within this layout by changing which tokens occupy the window or how their positions are encoded. The per-head KV layout itself, a dominant contributor to streaming memory and latency, has been mostly left unchanged. In this paper, we present the first study of Multi-Head Latent Attention (MLA) in video diffusion. VideoMLA replaces per-head keys and values with a shared low-rank content latent and a shared decoupled 3D-RoPE positional key, reducing per-token KV memory by 92.7% at every cached layer. We further investigate why MLA succeeds in video diffusion even though the spectral assumption often used to motivate it in language models does not hold: pretrained video attention is not low-rank, with 99%-energy effective rank far above any practical latent dimension. VideoMLA retains quality at compression ratios where direct spectral approximation would predict large reconstruction error. We show that the MLA bottleneck, rather than the pretrained spectrum, determines the effective rank: both spectral and random initialization occupy nearly the full rank budget from initialization, and training preserves this budget while adapting within it. On VBench, VideoMLA matches short-horizon streaming video diffusion baselines, achieves the best overall score at long horizons among evaluated methods, and improves throughput by 1.23x on a single B200.