Research map / Image generation
Fast diffusion sampling: research map
2,155 papers on Fast diffusion sampling in Image generation, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 4 clusters and 16 approaches. The busiest year so far is 2026.
What it is
Diffusion models generate by removing noise step by step, which is slow: early models needed hundreds of steps. This topic studies the theory of diffusion and flow matching and makes sampling fast through better solvers, guidance and distillation into few-step models.
Terms you will see on this page:
- Denoising diffusion
- generating by repeatedly denoising, starting from pure noise.
- Flow matching / rectified flow
- learning straight paths from noise to data so few steps suffice.
- Distillation
- training a fast few-step model to imitate a slow many-step one.
Start reading with the most cited papers here:
- Denoising Diffusion Probabilistic Models (NeurIPS 2020 · 5,610 citations)
- Diffusion Models Beat GANs on Image Synthesis (NeurIPS 2021 · 2,151 citations)
Within Image generation, its share grew from 22.1% in 2023–24 to 32.3% in 2025–26 (460 → 1,322 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore Fast diffusion sampling in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
What changed recently
Approaches whose share of the topic moved most between 2023–24 and 2025–26 (ICML, NeurIPS, ICLR, CVPR and AAAI):
- Masked diffusion language model decoding rising: 1% → 11% of the topic (4 → 142 papers)
- Flow matching distillation rising: 2% → 6% of the topic (9 → 85 papers)
- Training-free diffusion acceleration rising: 11% → 15% of the topic (51 → 203 papers)
- Diffusion models as perception tools falling: 17% → 8% of the topic (80 → 109 papers)
- Denoising diffusion probabilistic models falling: 8% → 4% of the topic (37 → 48 papers)
Papers from 2025–2026 that are already cited most:
- DiffusionDrive: Truncated Diffusion Model for End-to-End Autonomous Driving (CVPR 2025 · 87 citations)
- DepthFM: Fast Generative Monocular Depth Estimation with Flow Matching (AAAI 2025 · 37 citations)
- GoalFlow: Goal-Driven Flow Matching for Multimodal Trajectories Generation in End-to-End Autonomous Driving (CVPR 2025 · 30 citations)
- Mean Flows for One-step Generative Modeling (NeurIPS 2025 · 26 citations)
Approaches and key papers
Diffusion theory and guidance · 1,123 papers
Keywords: score, distribution, denoising diffusion
Approaches in this cluster:
- Diffusion theory and design space (269 papers)
Analyze convergence of score-based diffusion and optimize noise schedules and training design. - Distributional control of diffusion (134 papers)
Steer pretrained generators for diversity, minority sampling, and distribution-level guidance. - Denoising diffusion probabilistic models (113 papers)
Develop DDPM variants, implicit samplers, and bridge formulations for generation and denoising. - Guidance for conditional diffusion (130 papers)
Improve classifier-free and rectified guidance for conditional generation. - Diffusion posterior sampling (93 papers)
Solve inverse problems with diffusion priors through posterior sampling. - Diffusion theory and learned samplers (135 papers)
Provides variational and non-asymptotic views of diffusion and learns diffusions for sampling. - Diffusion models as perception tools (249 papers)
Use diffusion models for segmentation, adaptation, and data generation in discriminative tasks.
Most cited and most cited since 2024:
- Denoising Diffusion Probabilistic Models (NeurIPS 2020 · 5,610 citations)
- Diffusion Models Beat GANs on Image Synthesis (NeurIPS 2021 · 2,151 citations)
- Plug-In Diffusion Model for Sequential Recommendation (AAAI 2024 · 63 citations)
- Region-Disentangled Diffusion Model for High-Fidelity PPG-to-ECG Translation (AAAI 2024 · 39 citations)
Few-step and faster sampling · 582 papers
Keywords: steps, distillation, time
Approaches in this cluster:
- Training-free diffusion acceleration (316 papers)
Speed diffusion inference through caching, pruning, early exit, and adaptive step selection. - Fast diffusion ODE solvers (140 papers)
Reduce sampling steps with exponential integrators, parallel sampling, and optimized timesteps. - Distillation for few-step generation (99 papers)
Distill diffusion models and datasets into fast generators using distribution matching. - Accelerated diffusion policies (27 papers)
Speed robotic and driving diffusion policies with distillation and pruning.
Most cited and most cited since 2024:
- Parallel WaveNet: Fast High-Fidelity Speech Synthesis (ICML 2018 · 335 citations)
- Professor Forcing: A New Algorithm for Training Recurrent Networks (NeurIPS 2016 · 321 citations)
- One-step Diffusion with Distribution Matching Distillation (CVPR 2024 · 146 citations)
- DeepCache: Accelerating Diffusion Models for Free (CVPR 2024 · 136 citations)
Flow matching and rectified flow · 228 papers
Keywords: flow, matching, rectified
Approaches in this cluster:
- Flow matching distillation (113 papers)
Distill and align flow matching models into few-step generators. - Rectified flow straightening (77 papers)
Straighten flow trajectories with reflow to speed up generation. - Guidance and alignment for flow matching (38 papers)
Improve guidance, stability, and reward alignment in flow matching models.
Most cited and most cited since 2024:
- Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow (ICLR 2023 · 84 citations)
- FlowDiffuser: Advancing Optical Flow Estimation with Diffusion Models (CVPR 2024 · 42 citations)
- DepthFM: Fast Generative Monocular Depth Estimation with Flow Matching (AAAI 2025 · 37 citations)
- GoalFlow: Goal-Driven Flow Matching for Multimodal Trajectories Generation in End-to-End Autonomous Driving (CVPR 2025 · 30 citations)
Discrete and language diffusion · 222 papers
Keywords: discrete, diffusion language, masked
Approaches in this cluster:
- Masked diffusion language model decoding (152 papers)
Accelerate and improve parallel decoding of masked diffusion language models. - Discrete diffusion solvers (70 papers)
Develop fast solvers, correctors, and hybrid formulations for discrete-state diffusion.
Most cited and most cited since 2024:
- Structured Denoising Diffusion Models in Discrete State-Spaces (NeurIPS 2021 · 82 citations)
- Analog Bits: Generating Discrete Data using Diffusion Models with Self-Conditioning (ICLR 2023 · 80 citations)
- Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion (NeurIPS 2024 · 47 citations)
- Simplified and Generalized Masked Diffusion for Discrete Data (NeurIPS 2024 · 15 citations)
Related topics in Image generation
- Autoregressive and latent generation (2,100)
- GANs (780)
- Super-resolution (771)
- Editing and style transfer (1,430)
- Image restoration (2,009)
- Text-to-image diffusion (1,916)
