mlsanityData to 14 Oct 2026

Why Are Conditional Generative Models Better Than Unconditional Ones?

arXiv 2212.00362 ↗First posted 1 Dec 2022cs.LGcited by 28

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

Extensive empirical evidence demonstrates that conditional generative models are easier to train and perform better than unconditional ones by exploiting the labels of data. So do score-based diffusion models. In this paper, we analyze the phenomenon formally and identify that the key of conditional learning is to partition the data properly. Inspired by the analyses, we propose self-conditioned diffusion models (SCDM), which is trained conditioned on indices clustered by the k-means algorithm on the features extracted by a model pre-trained in a self-supervised manner. SCDM significantly improves the unconditional model across various datasets and achieves a record-breaking FID of 3.94 on ImageNet 64x64 without labels. Besides, SCDM achieves a slightly better FID than the corresponding conditional model on CIFAR10.

Key facts. Cited by 28 papers in the map (data through 14 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 16 of its 22 references that are in the map. First posted on arXiv 1 Dec 2022.

Builds on

16 of 22 references.

Closest ancestors · 6

Same direction, newest first

  1. 2022DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps2 Jun
  2. 2022Analytic-DPM: an Analytic Estimate of the Optimal Reverse Variance in Diffusion Probabilistic Models17 Jan
  3. 2021Diffusion Models Beat GANs on Image Synthesis11 May
  4. 2021Improved Denoising Diffusion Probabilistic Models18 Feb
  5. 2021Maximum Likelihood Training of Score-Based Diffusion Models22 Jan
  6. 2020Score-Based Generative Modeling through Stochastic Differential Equations26 Nov

Other references · 10

  1. 2014Adam: A Method for Stochastic Optimizationcited by 37,240
  2. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 12,049
  3. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,513
  4. 2014Deep Learning Face Attributes in the Wildcited by 3,693
  5. 2018Large Scale GAN Training for High Fidelity Natural Image Synthesiscited by 2,626
  6. 2020Improved Baselines with Momentum Contrastive Learningcited by 2,287
  7. 2021An Empirical Study of Training Self-Supervised Vision Transformerscited by 1,616
  8. 2021Instance-Conditioned GANcited by 121
  9. 2017Triple Generative Adversarial Netscited by 60
  10. 2020Self-labeled Conditional GANscited by 8

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2023: 7; 2024: 8; 2025: 10; 2026: 3.

28 papers.

Sort:

By area:

  1. 2024GenAD: Generalized Predictive Model for Autonomous Drivingcited by 150
  2. 2023Return of Unconditional Generation: A Self-supervised Representation Generation Methodcited by 98
  3. 2023SlotDiffusion: Object-Centric Generative Modeling with Diffusion Modelscited by 85
  4. 2024A Survey on Diffusion Models for Time Series and Spatio-Temporal Datacited by 67
  5. 2025Proteina: Scaling Flow-based Protein Structure Generative Modelscited by 65
  6. 2024Trajectory Consistency Distillation: Improved Latent Consistency Distillation by Semi-Linear Consistency Function with Trajectory Mappingcited by 46
  7. 2023Diffusion Models and Semi-Supervised Learners Benefit Mutually with Few Labelscited by 37
  8. 2023Toward Understanding Generative Data Augmentationcited by 25
  9. 2024DisCo-Diff: Enhancing Continuous Diffusion Models with Discrete Latentscited by 20
  10. 2025Highly Compressed Tokenizer Can Generate Without Trainingcited by 18
  11. 2023Visual Chain-of-Thought Diffusion Modelscited by 12
  12. 2024MotionMix: Weakly-Supervised Diffusion for Controllable Motion Generationcited by 10
16 more
  1. 2025What Happens Next? Anticipating Future Motion by Generating Point Trajectoriescited by 10
  2. 2023Guided Diffusion from Self-Supervised Diffusion Featurescited by 9
  3. 2024Cross-Domain Graph Data Scaling: A Showcase with Diffusion Modelscited by 8
  4. 2025Versatile Cardiovascular Signal Generation with a Unified Diffusion Transformercited by 7
  5. 2024Rethinking cluster-conditioned diffusion models for label-free image synthesiscited by 5
  6. 2023Masked Generative Modeling with Enhanced Sampling Schemecited by 3
  7. 2025Conditioning Residuals for Diffusion Models via Representation Feedbackcited by 3
  8. 2025Compositional Discrete Latent Code for High Fidelity, Productive Diffusion Modelscited by 3
  9. 2025Rethinking Diffusion Model in High Dimensioncited by 2
  10. 2024Scaling Up Deep Clustering Methods Beyond ImageNet-1Kcited by 1
  11. 2025Leveraging Discriminative Latent Representations for Conditioning GAN-Based Speech Enhancementcited by 1
  12. 2026A Hybrid Conditional Diffusion-DeepONet Framework for High-Fidelity Stress Prediction in Hyperelastic Materialscited by 1
  13. 2025CAFusion: Controllable Anatomical Synthesis of Perirectal Lymph Nodes via SDF-guided Diffusioncited by 0
  14. 2025Challenges and Limitations of Generative AI in Synthesizing Wearable Sensor Datacited by 0
  15. 2026Flow Matching with Optimized Subclass Priors for Medical Image Augmentationcited by 0
  16. 2026From Pixels, Without Pre-training: Joint Generative and Self-Supervised Representation Learning in One Modelcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2023Conditional Generation from Unconditional Diffusion Models using Denoiser Representations
  2. 2021Conditional Image Generation with Score-Based Diffusion Models
  3. 2024On Improved Conditioning Mechanisms and Pre-training Strategies for Diffusion Models
  4. 2023Zero-Shot Conditioning of Score-Based Diffusion Models by Neuro-Symbolic Constraints
  5. 2022Elucidating the Design Space of Diffusion-Based Generative Models
  6. 2023Conditional Image Generation with Pretrained Generative Model
  7. 2024Unveil Conditional Diffusion Models with Classifier-free Guidance: A Sharp Statistical Theory
  8. 2024Improving Training-free Conditional Diffusion Model via Fisher Information