mlsanityData to 14 Oct 2026

Unsupervised Decomposition and Recombination with Discriminator-Driven Diffusion Models

arXiv 2601.22057 ↗First posted 29 Jan 2026cs.CV cs.AIcited by 3

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

Decomposing complex data into factorized representations can reveal reusable components and enable synthesizing new samples via component recombination. We investigate this in the context of diffusion-based models that learn factorized latent spaces without factor-level supervision. In images, factors can capture background, illumination, and object attributes; in robotic videos, they can capture reusable motion components. To improve both latent factor discovery and quality of compositional generation, we introduce an adversarial training signal via a discriminator trained to distinguish between single-source samples and those generated by recombining factors across sources. By optimizing the generator to fool this discriminator, we encourage physical and semantic consistency in the resulting recombinations. Our method outperforms implementations of prior baselines on CelebA-HQ, Virtual KITTI, CLEVR, and Falcor3D, achieving lower FID scores and better disentanglement as measured by MIG and MCC. Furthermore, we demonstrate a novel application to robotic video trajectories: by recombining learned action components, we generate diverse sequences that significantly increase state-space coverage for exploration on the LIBERO benchmark.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 53 of its 70 references that are in the map. First posted on arXiv 29 Jan 2026.

Builds on

53 of 70 references.

Closest ancestors · 6

Same direction, newest first

  1. 2024Compositional Image Decomposition with Diffusion Models27 Jun
  2. 2023Learning Energy-Based Models by Cooperative Diffusion Recovery Likelihood10 Sep
  3. 2023Universal Guidance for Diffusion Models14 Feb
  4. 2022Diffusion models as plug-and-play priors17 Jun
  5. 2020Denoising Diffusion Implicit Models6 Oct
  6. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 47

  1. 2017Progressive Growing of GANs for Improved Quality, Stability, and Variationcited by 2,988
  2. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,838
  3. 2022InstructPix2Pix: Learning to Follow Image Editing Instructionscited by 2,611
  4. 2016CLEVR: A Diagnostic Dataset for Compositional Language and Elementary Visual Reasoningcited by 1,457
  5. 2023LIBERO: Benchmarking Knowledge Transfer for Lifelong Robot Learningcited by 1,410
  6. 2022Planning with Diffusion for Flexible Behavior Synthesiscited by 1,061
  7. 2023ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generationcited by 1,045
  8. 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 941
  9. 2016InfoGAN: Interpretable Representation Learning by Information Maximizing Generative Adversarial Netscited by 860
  10. 2024RDT-1B: a Diffusion Foundation Model for Bimanual Manipulationcited by 691
  11. 2022Compositional Visual Generation with Composable Diffusion Modelscited by 624
  12. 2018Disentangling by Factorisingcited by 616
35 more
  1. 2023Learning Universal Policies via Text-Guided Video Generationcited by 522
  2. 2022Is Conditional Generative Modeling all you need for Decision-Making?cited by 516
  3. 2019Variational Autoencoders and Nonlinear ICA: A Unifying Frameworkcited by 503
  4. 2023Directly Fine-Tuning Diffusion Models on Differentiable Rewardscited by 448
  5. 2019MONet: Unsupervised Scene Decomposition and Representationcited by 392
  6. 2016Virtual Worlds as Proxy for Multi-Object Tracking Analysiscited by 341
  7. 2021How to Train Your Energy-Based Modelscited by 274
  8. 2016Unsupervised Feature Extraction by Time-Contrastive Learning and Nonlinear ICAcited by 273
  9. 2023Zero-Shot Robotic Manipulation with Pretrained Image-Editing Diffusion Modelscited by 269
  10. 2018Stochastic Video Generation with a Learned Priorcited by 245
  11. 2020Swapping Autoencoder for Deep Image Manipulationcited by 215
  12. 2024RoboDreamer: Learning Compositional World Models for Robot Imaginationcited by 195
  13. 2023Learning to Act from Actionless Videos through Dense Correspondencescited by 164
  14. 2021GENESIS-V2: Inferring Unordered Object Representations without Iterative Refinementcited by 109
  15. 2023SlotDiffusion: Object-Centric Generative Modeling with Diffusion Modelscited by 85
  16. 2022Linear Causal Disentanglement via Interventionscited by 84
  17. 2024Fine-Tuning Discrete Diffusion Models via Reward Optimization with Applications to DNA and Protein Designcited by 79
  18. 2021Learning to Compose Visual Relationscited by 72
  19. 2021Unsupervised Learning of Compositional Energy Conceptscited by 71
  20. 2023Compositional Generalization from First Principlescited by 64
  21. 2016Generative Multi-Adversarial Networkscited by 60
  22. 2023Additive Decoders for Latent Variables Identification and Cartesian-Product Extrapolationcited by 58
  23. 2019Unsupervised Model Selection for Variational Disentangled Representation Learningcited by 46
  24. 2023AdvDiff: Generating Unrestricted Adversarial Examples using Diffusion Modelscited by 46
  25. 2022UVCGAN: UNet Vision Transformer cycle-consistent GAN for unpaired image-to-image translationcited by 42
  26. 2025Fine-Tuning Discrete Diffusion Models with Policy Gradient Methodscited by 38
  27. 2022VAEL: Bridging Variational Autoencoders and Probabilistic Logic Programmingcited by 31
  28. 2024Grounding Video Models to Actions through Goal Conditioned Explorationcited by 26
  29. 2022Online Adaptive Policy Selection in Time-Varying Systems: No-Regret via Contractive Perturbationscited by 14
  30. 2023Object-Centric Learning with Slot Mixture Modulecited by 11
  31. 2024Mean-Field Sampling for Cooperative Multi-Agent Reinforcement Learningcited by 9
  32. 2024Online Policy Optimization in Unknown Nonlinear Systemscited by 8
  33. 2024Peer-to-Peer Learning Dynamics of Wide Neural Networkscited by 8
  34. 2024Few-Shot Task Learning through Inverse Generative Modelingcited by 4
  35. 2025Feel-Good Thompson Sampling for Contextual Bandits: a Markov Chain Monte Carlo Showdowncited by 4

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 3.

3 papers.

  1. 2026Continuous Latent Contexts Enable Efficient Online Learning in Transformerscited by 3
  2. 2026Learning Approximate Nash Equilibria in Cooperative Multi-Agent Reinforcement Learning via Mean-Field Subsamplingcited by 2
  3. 2026Graphon Mean-Field Subsampling for Cooperative Heterogeneous Multi-Agent Reinforcement Learningcited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2026Factor-Aware Mixture-of-Experts with Pretrained Encoder for Combinatorial Generalization
  2. 2026SlotDiT: Object-Centric Representations for Diffusion Transformers
  3. 2025Disentanglement Beyond Static vs. Dynamic: A Benchmark and Evaluation Framework for Multi-Factor Sequential Representations
  4. 2026Diffusing in the Right Space: A Systematic Study of Latent Diffusability
  5. 2023Video Probabilistic Diffusion Models in Projected Latent Space
  6. 2025From Pixels to Factors: Learning Independently Controllable State Variables for Reinforcement Learning
  7. 2026Diff-CA: Separating Common and Salient Factors with Diffusion Models
  8. 2025Masked Autoencoders Are Effective Tokenizers for Diffusion Models