# Where Does Generative Difficulty Reside? An Empirical Study of Target Representations

arXiv: https://arxiv.org/abs/2608.00626 · HTML: https://mlsanity.com/paper/2608.00626/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 28 of its 39 references that are in the map.
- First posted on arXiv 1 Aug 2026.

## Builds on (most cited first, 28 of 28)

- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [Score-Based Generative Modeling through Stochastic Differential Equations](https://mlsanity.com/paper/2011.13456/) (2020)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Masked Autoencoders Are Scalable Vision Learners](https://mlsanity.com/paper/2111.06377/) (2021)
- [Deep Unsupervised Learning using Nonequilibrium Thermodynamics](https://mlsanity.com/paper/1503.03585/) (2015)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [Zero-Shot Text-to-Image Generation](https://mlsanity.com/paper/2102.12092/) (2021)
- [Scaling Rectified Flow Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2403.03206/) (2024)
- [Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow](https://mlsanity.com/paper/2209.03003/) (2022)
- [Taming Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2012.09841/) (2020)
- [Improved Techniques for Training GANs](https://mlsanity.com/paper/1606.03498/) (2016)
- [MaskGIT: Masked Generative Image Transformer](https://mlsanity.com/paper/2202.04200/) (2022)
- [Conditional Image Generation with PixelCNN Decoders](https://mlsanity.com/paper/1606.05328/) (2016)
- [SiT: Exploring Flow and Diffusion-based Generative Models with Scalable Interpolant Transformers](https://mlsanity.com/paper/2401.08740/) (2024)
- [Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Think](https://mlsanity.com/paper/2410.06940/) (2024)
- [Autoregressive Image Generation without Vector Quantization](https://mlsanity.com/paper/2406.11838/) (2024)
- [Diffusion Transformers with Representation Autoencoders](https://mlsanity.com/paper/2510.11690/) (2025)
- [MAGE: MAsked Generative Encoder to Unify Representation Learning and Image Synthesis](https://mlsanity.com/paper/2211.09117/) (2022)
- [PixelDiT: Pixel Diffusion Transformers for Image Generation](https://mlsanity.com/paper/2511.20645/) (2025)
- [One-step Latent-free Image Generation with Pixel Mean Flows](https://mlsanity.com/paper/2601.22158/) (2026)
- [Latent Forcing: Reordering the Diffusion Trajectory for Pixel-Space Image Generation](https://mlsanity.com/paper/2602.11401/) (2026)
- [Self-Supervised Flow Matching for Scalable Multi-Modal Synthesis](https://mlsanity.com/paper/2603.06507/) (2026)
- [Improved Baselines with Representation Autoencoders](https://mlsanity.com/paper/2605.18324/) (2026)
- [Effective and Efficient Masked Image Generation Models](https://mlsanity.com/paper/2503.07197/) (2025)
- [Trajectory Forcing: Structure-First Generation with Controllable Semantic Trajectories](https://mlsanity.com/paper/2606.22527/) (2026)

## Abstract (arXiv)

The target representation defines the distribution an image generator must learn, yet it is often treated as an interchangeable interface. This assumption is particularly questionable for continuous masked generators, which combine contextual inference from visible tokens with conditional modeling of each missing token. We study raw pixels, SD-VAE latents and DINOv2 as well as MAE representation-autoencoder features within a unified masked autoregressive rectified-flow model. Under a shared ImageNet training budget, these spaces exhibit distinct optimization and inference regimes. DINOv2 converges fastest in both iterations and computation but benefits strongly from a wider local denoiser and direct context fusion. Pixels optimize substantially more slowly and require a different prediction, masking, and guidance configuration. MAE reconstructs images more faithfully and exhibits clear semantic clustering, yet produces generations substantially worse than DINOv2. The representations also respond differently to classifier-free guidance and occupy distinct precision-recall trade-offs. Together, our results show that compression, reconstruction fidelity, token dimensionality, and visible semantic clustering do not individually predict generative behavior. Instead, target representations redistribute difficulty across contextual modeling, per-token denoising, and inference-time distributional control.