mlsanityData to 14 Oct 2026

HybridStitch: Pixel and Timestep Level Model Stitching for Diffusion Acceleration

arXiv 2603.07815 ↗First posted 8 Mar 2026cs.CV cs.AIcited by 2

On the map: Diffusion models → Image generation, text-to-image and editing

Abstract (arXiv)

Iterative diffusion generation requires different amounts of model capacity at different token positions. Prior model combinations allocate capacity by denoising step or by drafting and verification role. We present HybridStitch, a training-free framework that uses token-level feedback to select the model supplying each adopted update. Continuous image refinement and discrete language commitment operate on a shared generation state, while model-specific KV buffers provide attention context for selected queries. HybridStitch achieves 1.63--1.68× faster image generation and 1.15--1.58× diffusion large language model (dLLM) throughput on four main benchmarks. Estimated dLLM decode query cost decreases by 50--66% with comparable generation quality.

Key facts. Cited by 2 papers in the map (data through 14 Oct 2026). Placed in Image generation, text-to-image and editing, one of 14 directions in Diffusion models. Builds on 30 of its 39 references that are in the map. First posted on arXiv 8 Mar 2026.

Builds on

30 of 39 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025FLUX.1 Kontext: Flow Matching for In-Context Image Generation and Editing in Latent Space17 Jun
  2. 2024DiffSim: Taming Diffusion Models for Evaluating Visual Similarity19 Dec
  3. 2023TokenCompose: Text-to-Image Diffusion with Token-level Supervision6 Dec

Other references · 27

  1. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,863
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  3. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  4. 2023SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesiscited by 4,356
  5. 2024Scaling Rectified Flow Transformers for High-Resolution Image Synthesiscited by 4,181
  6. 2021CLIPScore: A Reference-free Evaluation Metric for Image Captioningcited by 2,424
  7. 2025Z-Image: An Efficient Image Generation Foundation Model with Single-Stream Diffusion Transformercited by 277
  8. 2025Sparse VideoGen: Accelerating Video Diffusion Transformers with Spatial-Temporal Sparsitycited by 217
  9. 2024SageAttention: Accurate 8-Bit Attention for Plug-and-play Inference Accelerationcited by 160
  10. 2024Learning-to-Cache: Accelerating Diffusion Transformer via Layer Cachingcited by 142
  11. 2025From Reusing to Forecasting: Accelerating Diffusion Models with TaylorSeerscited by 136
  12. 2024SageAttention2: Efficient Attention with Thorough Outlier Smoothing and Per-thread INT4 Quantizationcited by 114
15 more
  1. 2024DistriFusion: Distributed Parallel Inference for High-Resolution Diffusion Modelscited by 98
  2. 2025SageAttention3: Microscaling FP4 Attention for Inference and An Exploration of 8-Bit Trainingcited by 66
  3. 2025SLA: Beyond Sparsity in Diffusion Transformers via Fine-Tunable Sparse-Linear Attentioncited by 58
  4. 2024xDiT: an Inference Engine for Diffusion Transformers (DiTs) with Massive Parallelismcited by 56
  5. 2024Adversarial Diffusion Compression for Real-World Image Super-Resolutioncited by 52
  6. 2025Distribution Matching Distillation Meets Reinforcement Learningcited by 50
  7. 2025Decoupled DMD: CFG Augmentation as the Spear, Distribution Matching as the Shieldcited by 45
  8. 2024T-Stitch: Accelerating Sampling in Pre-Trained Diffusion Models with Trajectory Stitchingcited by 33
  9. 2026SLA2: Sparse-Linear Attention with Learnable Routing and QATcited by 22
  10. 2026SpargeAttention2: Trainable Sparse Attention via Hybrid Top-k+Top-p Masking and Distillation Fine-Tuningcited by 18
  11. 2025Dual Prompting Image Restoration with Diffusion Transformerscited by 15
  12. 2025MoDM: Efficient Serving for Image Generation via Mixture-of-Diffusion Modelscited by 12
  13. 2025SRDiffusion: Accelerate Video Diffusion Inference via Sketching-Rendering Cooperationcited by 7
  14. 2024FlexCache: Flexible Approximate Cache System for Video Diffusioncited by 5
  15. 2025Fewer Denoising Steps or Cheaper Per-Step Inference: Towards Compute-Optimal Diffusion Model Deploymentcited by 5

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026CachedSearch: Training-Free Cached Exploration for Test-Time Search in Video Diffusioncited by 3
  2. 2026Speculative Decoding for Autoregressive Video Generationcited by 2

Similar, not cited

Close in wording, no citation either way

  1. 2024Diffusion Model Compression for Image-to-Image Translation
  2. 2024Controllable Generation with Text-to-Image Diffusion Models: A Survey
  3. 2026TMPDiff: Temporal Mixed-Precision for Diffusion Models
  4. 2023Accelerating Text-to-Image Editing via Cache-Enabled Sparse Diffusion Inference
  5. 2026Toward Early Quality Assessment of Text-to-Image Diffusion Models
  6. 2024MaxFusion: Plug&Play Multi-Modal Generation in Text-to-Image Diffusion Models
  7. 2023Approximate Caching for Efficiently Serving Diffusion Models
  8. 2024DiffEditor: Boosting Accuracy and Flexibility on Diffusion-based Image Editing