mlsanityData to 14 Oct 2026

MeanCache: From Instantaneous to Average Velocity for Accelerating Flow Matching Inference

arXiv 2601.19961 ↗First posted 27 Jan 2026cs.LG cs.AI cs.CVcited by 10

On the map: Diffusion models → Flow matching, time series and science

Abstract (arXiv)

We present MeanCache, a training-free caching framework for efficient Flow Matching inference. Existing caching methods reduce redundant computation but typically rely on instantaneous velocity information (e.g., feature caching), which often leads to severe trajectory deviations and error accumulation under high acceleration ratios. MeanCache introduces an average-velocity perspective: by leveraging cached Jacobian--vector products (JVP) to construct interval average velocities from instantaneous velocities, it effectively mitigates local error accumulation. To further improve cache timing and JVP reuse stability, we develop a trajectory-stability scheduling strategy as a practical tool, employing a Peak-Suppressed Shortest Path under budget constraints to determine the schedule. Experiments on FLUX.1, Qwen-Image, and HunyuanVideo demonstrate that MeanCache achieves 4.12X and 4.56X and 3.59X acceleration, respectively, while consistently outperforming state-of-the-art caching baselines in generation quality. We believe this simple yet effective approach provides a new perspective for Flow Matching inference and will inspire further exploration of stability-driven acceleration in commercial-scale generative models.

Key facts. Cited by 10 papers in the map (data through 14 Oct 2026). Placed in Flow matching, time series and science, one of 14 directions in Diffusion models. Builds on 55 of its 66 references that are in the map. First posted on arXiv 27 Jan 2026.

Builds on

55 of 66 references.

Closest ancestors · 1

Same direction, newest first

  1. 2022Flow Matching for Generative Modeling6 Oct

Other references · 54

  1. 2020Denoising Diffusion Probabilistic Modelscited by 21,076
  2. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  3. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  4. 2020Denoising Diffusion Implicit Modelscited by 9,781
  5. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 9,122
  6. 2022Photorealistic Text-to-Image Diffusion Models with Deep Language Understandingcited by 6,355
  7. 2022Scalable Diffusion Models with Transformerscited by 6,054
  8. 2022FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awarenesscited by 3,673
  9. 2022Elucidating the Design Space of Diffusion-Based Generative Modelscited by 3,045
  10. 2021CLIPScore: A Reference-free Evaluation Metric for Image Captioningcited by 2,424
  11. 2015Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Codingcited by 2,329
  12. 2015Variational Inference with Normalizing Flowscited by 2,170
38 more
  1. 2023Consistency Modelscited by 1,822
  2. 2023VBench: Comprehensive Benchmark Suite for Video Generative Modelscited by 1,822
  3. 2024HunyuanVideo: A Systematic Framework For Large Video Generative Modelscited by 1,660
  4. 2023ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generationcited by 1,045
  5. 2023One-step Diffusion with Distribution Matching Distillationcited by 1,019
  6. 2022Building Normalizing Flows with Stochastic Interpolantscited by 995
  7. 2022DPM-Solver++: Fast Solver for Guided Sampling of Diffusion Probabilistic Modelscited by 938
  8. 2023Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inferencecited by 846
  9. 2023Adversarial Diffusion Distillationcited by 759
  10. 2024Improved Distribution Matching Distillation for Fast Image Synthesiscited by 684
  11. 2024Open-Sora: Democratizing Efficient Video Production for Allcited by 652
  12. 2023LLM-Pruner: On the Structural Pruning of Large Language Modelscited by 650
  13. 2025Mean Flows for One-step Generative Modelingcited by 561
  14. 2023Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusioncited by 459
  15. 2024One Step Diffusion via Shortcut Modelscited by 391
  16. 2023DeepCache: Accelerating Diffusion Models for Freecited by 359
  17. 2024Fast High-Resolution Image Synthesis with Latent Adversarial Diffusion Distillationcited by 279
  18. 2024Timestep Embedding Tells: It's Time to Cache for Video Diffusion Modelcited by 269
  19. 2022Post-training Quantization on Diffusion Modelscited by 267
  20. 2021Gotta Go Fast When Generating Data with Score-Based Modelscited by 219
  21. 2024T2V-CompBench: A Comprehensive Benchmark for Compositional Text-to-video Generationcited by 185
  22. 2024Delta-DiT: A Training-Free Acceleration Method Tailored for Diffusion Transformerscited by 175
  23. 2024Accelerating Diffusion Transformers with Token-wise Feature Cachingcited by 156
  24. 2023Cache Me if You Can: Accelerating Diffusion Models through Block Cachingcited by 151
  25. 2024Real-Time Video Generation with Pyramid Attention Broadcastcited by 149
  26. 2024Consistency Models Made Easycited by 144
  27. 2024Learning-to-Cache: Accelerating Diffusion Transformer via Layer Cachingcited by 142
  28. 2024FORA: Fast-Forward Caching in Diffusion Transformer Accelerationcited by 137
  29. 2025From Reusing to Forecasting: Accelerating Diffusion Models with TaylorSeerscited by 136
  30. 2025Inductive Moment Matchingcited by 113
  31. 2024DistriFusion: Distributed Parallel Inference for High-Resolution Diffusion Modelscited by 98
  32. 2023Temporal Dynamic Quantization for Diffusion Modelscited by 81
  33. 2024xDiT: an Inference Engine for Diffusion Transformers (DiTs) with Massive Parallelismcited by 56
  34. 2023OMS-DPM: Optimizing the Model Schedule for Diffusion Probabilistic Modelscited by 45
  35. 2025DiCache: Let Diffusion Model Determine Its Own Cachecited by 42
  36. 2024AsyncDiff: Parallelizing Diffusion Models by Asynchronous Denoisingcited by 22
  37. 2024DSP: Dynamic Sequence Parallelism for Multi-Dimensional Transformerscited by 15
  38. 2018Deep Global-Relative Networks for End-to-End 6-DoF Visual Localization and Odometrycited by 10

50 most cited of 54 All 54 (JSON)

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 10.

10 papers.

  1. 2026Chain-of-Trajectories: Unlocking the Intrinsic Generative Optimality of Diffusion Models via Graph-Theoretic Planningcited by 1
  2. 2026OTCache: Optimal Transport for Geometry-Aware Caching in Diffusion Modelscited by 1
  3. 2026BAG: Budget-Aware Gating for Diffusion Cachingcited by 1
  4. 2026LayerCache: Exploiting Layer-wise Velocity Heterogeneity for Efficient Flow Matching Inferencecited by 0
  5. 2026MediaClaw: Multimodal Intelligent-Agent Platform Technical Reportcited by 0
  6. 2026Token-Operations-Oriented Inference Optimization Techniques for Large Modelscited by 0
  7. 2026Tensor Field Modelscited by 0
  8. 2026The Golden Path Hypothesis: Reusable Schedules in Diffusion Cachingcited by 0
  9. 2026TRAC: Trajectory-aware Reuse and Adaptive Correction for Efficient Autoregressive Video Generationcited by 0
  10. 2026Rethinking Prediction Horizons: Decoupling Model Queries from State Updates in Few-Step MeanFlowcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Accelerating Rectified Flow Models via Trajectory-Aware Caching
  2. 2026ReAL: Accelerating Flow Matching through Segment Advancement with Shared Lookahead
  3. 2026Stable Velocity: A Variance Perspective on Flow Matching
  4. 2026The Velocity Deficit: Initial Energy Injection for Flow Matching
  5. 2025Graph Flow Matching: Enhancing Image Generation with Neighbor-Aware Flow Fields
  6. 2026Mean Flow Distillation: Robust and Stable Distillation for Flow Matching Models
  7. 2026Exploring and Exploiting Stability in Latent Flow Matching
  8. 2025Blockwise Flow Matching: Improving Flow Matching Models For Efficient High-Quality Generation