Diffusion & Flow
Practice Diffusion & Flow in Python with research-based exercises, implementation guides, and tests on PaperLeet.
- 6. DDPM
Build a generative model from a forward noising process and a learned reverse denoising process.
NeurIPS 2020 · 2 exercises - 8. DDIM
Use a non-Markovian sampling update to skip timesteps, with eta controlling stochasticity.
ICLR 2021 · 1 exercises - 7. Progressive Distillation
Distill two deterministic teacher steps into one student step, progressively reducing the sampling budget.
ICLR 2022 · 2 exercises - 9. Flow Matching
Learn a continuous generative flow by regressing the velocity field of conditional probability paths.
ICLR 2023 · 1 exercises - 20. Consistency Models
Map noisy states along the same generative trajectory to a consistent clean endpoint.
ICML 2023 · 1 exercises - 18. DMD
Use the score difference between real and generated distributions to construct a distribution-matching generator update.
CVPR 2024 · 1 exercises - 19. DMD2
Remove DMD's paired-regression requirement using faster fake-score updates and adversarial auxiliary training.
arXiv 2024 · 1 exercises - 15. MeanFlow
Learn an interval flow using the identity between instantaneous and average velocities for one-step generation.
arXiv 2025 · 1 exercises - 16. JiT
Predict clean data directly and explicitly convert between prediction and loss spaces.
arXiv 2025 · 1 exercises - 29. REPA
Align intermediate generative features with a frozen semantic teacher.
ICLR 2025 Oral · 1 exercises - 32. Block Diffusion
Combine autoregression across blocks with diffusion within blocks.
ICLR 2025 Oral · 1 exercises - 35. SVDQuant
Absorb difficult-to-quantize components in a high-precision low-rank branch and quantize the residual.
ICLR 2025 Spotlight · 1 exercises - 37. DAPS
Decouple measurement-consistency updates in clean space from renoising.
CVPR 2025 Oral · 1 exercises - 38. DiffusionNFT
Construct positive and negative implicit velocity fields from feedback to optimize diffusion models.
ICLR 2026 Oral · 1 exercises - 40. FD-Loss
Estimate a larger feature population from historical statistics while backpropagating Fréchet distance through the current batch only.
arXiv 2026 · 2 exercises