mlsanityData to 7 Oct 2026

Unified Latents (UL): How to train your latents

arXiv 2602.17270 ↗First posted 19 Feb 2026cs.LG cs.CVcited by 19

On the map: Diffusion models → Diffusion foundations and sampling

Abstract (arXiv)

We present Unified Latents (UL), a framework for learning latent representations that are jointly regularized by a diffusion prior and decoded by a diffusion model. By linking the encoder's output noise to the prior's minimum noise level, we obtain a simple training objective that provides a tight upper bound on the latent bitrate. On ImageNet-512, our approach achieves competitive FID of 1.4, with high reconstruction quality (PSNR) while requiring fewer training FLOPs than models trained on Stable Diffusion latents. On Kinetics-600, we set a new state-of-the-art FVD of 1.3.

Key facts. Cited by 19 papers in the map (data through 7 Oct 2026). Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models. Builds on 16 of its 17 references that are in the map. First posted on arXiv 19 Feb 2026.

Builds on

16 of 17 references.

Closest ancestors · 12

Same direction, newest first

  1. 2025Latent Diffusion Model without Variational Autoencoder17 Oct
  2. 2025Diffusion Transformers with Representation Autoencoders13 Oct
  3. 2024Simpler Diffusion (SiD2): 1.5 FID on ImageNet512 with pixel-space diffusion25 Oct
  4. 2024Deep Compression Autoencoder for Efficient High-Resolution Diffusion Models14 Oct
  5. 2022DiVAE: Photorealistic Images Synthesis with Denoising Diffusion Decoder1 Jun
  6. 2022DiffuseVAE: Efficient, Controllable and High-Fidelity Generation from Low-Dimensional Latents2 Jan
  7. 2021High-Resolution Image Synthesis with Latent Diffusion Models20 Dec
  8. 2021Diffusion Autoencoders: Toward a Meaningful and Decodable Representation30 Nov
  9. 2021Variational Diffusion Models1 Jul
  10. 2021Score-based Generative Modeling in Latent Space10 Jun
  11. 2021Cascaded Diffusion Models for High Fidelity Image Generation30 May
  12. 2020Denoising Diffusion Probabilistic Models19 Jun

Other references · 4

  1. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,893
  2. 2024An Image is Worth 32 Tokens for Reconstruction and Generationcited by 295
  3. 2019Preventing Posterior Collapse with delta-VAEscited by 86
  4. 2024Sample what you cant compresscited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 19.

19 papers.

Sort:

By area:

  1. 2026Improved Baselines with Representation Autoencoderscited by 24
  2. 2026Making Reconstruction FID Predictive of Diffusion Generation FIDcited by 12
  3. 2026What Matters for Diffusion-Friendly Latent Manifold? Prior-Aligned Autoencoders for Latent Diffusioncited by 7
  4. 2026PiD: Fast and High-Resolution Latent Decoding with Pixel Diffusioncited by 6
  5. 2026Lens: Rethinking Training Efficiency for Foundational Text-to-Image Modelscited by 4
  6. 2026WavCube: Unifying Speech Representation for Understanding and Generation via Semantic-Acoustic Joint Modelingcited by 3
  7. 2026Explorative Modeling: Unlocking a Third Pretraining Axis and End-to-End Generationcited by 3
  8. 2026SAME: A Semantically-Aligned Music Autoencodercited by 2
  9. 2026CrossFlow: One-Step Generation Across Latent and Pixel Spacescited by 2
  10. 2026WaiT for the Signal: Simple Frequency-Aware Flow-Matchingcited by 2
  11. 2026GenFirst: Generation Before Reconstruction for Stable End-to-End Latent Generative Modelingcited by 2
  12. 2026MUNI: Multimodal Unified Latent Diffusion for Coherent Any-to-Any Generationcited by 1
7 more
  1. 2026FlashDecoder: Real-Time Latent-to-Pixel Streaming Decoder with Transformerscited by 1
  2. 2026Understanding Latent Diffusability via Fisher Geometrycited by 0
  3. 2026Provably Learning Diffusion Models under the Manifold Hypothesis: Collapse and Refinecited by 0
  4. 2026CenSynCMB: Centre Maps and Physics-Guided Synthesis for Microbleed Detectioncited by 0
  5. 2026Weakly supervised concept Bottleneck Learning for Robust Two stage Object centric visual reasoningcited by 0
  6. 2026UNITE-AUDIO: Joint Learning of Continuous Tokenization and Latent Flow Matching for Text-to-Audio Generationcited by 0
  7. 2026MUNITE: Unified Multimodal Latent Inference for Any-to-Any Multimodal Generationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026End-to-End Training for Unified Tokenization and Latent Denoising
  2. 2025REGLUE Your Latents with Global and Local Semantics for Entangled Diffusion
  3. 2025DC-AE 1.5: Accelerating Diffusion Model Convergence with Structured Latent Space
  4. 2025USP: Unified Self-Supervised Pretraining for Image Generation and Understanding
  5. 2026DA-VAE: Plug-in Latent Compression for Diffusion via Detail Alignment
  6. 2026LDM-is-AE: Latent Diffusion Model is an Auto-Encoder for End-to-End Image Generation
  7. 2025OSCAR: One-Step Diffusion Codec Across Multiple Bit-rates
  8. 2026Addressing Detail Bottlenecks in Latent Diffusion for RGB-to-SWIR Image Translation