# How Neural Losses Shape VAE Latents

arXiv: https://arxiv.org/abs/2606.00635 · HTML: https://mlsanity.com/paper/2606.00635/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 25 of its 35 references that are in the map.
- First posted on arXiv 30 May 2026.

## Builds on (most cited first, 25 of 25)

- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [The Unreasonable Effectiveness of Deep Features as a Perceptual Metric](https://mlsanity.com/paper/1801.03924/) (2018)
- [DINOv2: Learning Robust Visual Features without Supervision](https://mlsanity.com/paper/2304.07193/) (2023)
- [Image-to-Image Translation with Conditional Adversarial Networks](https://mlsanity.com/paper/1611.07004/) (2016)
- [Deep Learning Face Attributes in the Wild](https://mlsanity.com/paper/1411.7766/) (2014)
- [Perceptual Losses for Real-Time Style Transfer and Super-Resolution](https://mlsanity.com/paper/1603.08155/) (2016)
- [Spectral Normalization for Generative Adversarial Networks](https://mlsanity.com/paper/1802.05957/) (2018)
- [Representation Alignment for Generation: Training Diffusion Transformers Is Easier Than You Think](https://mlsanity.com/paper/2410.06940/) (2024)
- [High Fidelity Neural Audio Compression](https://mlsanity.com/paper/2210.13438/) (2022)
- [The Perception-Distortion Tradeoff](https://mlsanity.com/paper/1711.06077/) (2017)
- [Autoencoding beyond pixels using a learned similarity metric](https://mlsanity.com/paper/1512.09300/) (2015)
- [Reconstruction vs. Generation: Taming Optimization Dilemma in Latent Diffusion Models](https://mlsanity.com/paper/2501.01423/) (2025)
- [High-Fidelity Generative Image Compression](https://mlsanity.com/paper/2006.09965/) (2020)
- [Understanding disentangling in β-VAE](https://mlsanity.com/paper/1804.03599/) (2018)
- [Variational Inference of Disentangled Latent Concepts from Unlabeled Observations](https://mlsanity.com/paper/1711.00848/) (2017)
- [REPA-E: Unlocking VAE for End-to-End Tuning with Latent Diffusion Transformers](https://mlsanity.com/paper/2504.10483/) (2025)
- [Rethinking Lossy Compression: The Rate-Distortion-Perception Tradeoff](https://mlsanity.com/paper/1901.07821/) (2019)
- [Variational Lossy Autoencoder](https://mlsanity.com/paper/1611.02731/) (2016)
- [Masked Autoencoders Are Effective Tokenizers for Diffusion Models](https://mlsanity.com/paper/2502.03444/) (2025)
- [EQ-VAE: Equivariance Regularized Latent Space for Improved Generative Image Modeling](https://mlsanity.com/paper/2502.09509/) (2025)
- [Taming VAEs](https://mlsanity.com/paper/1810.00597/) (2018)
- [Improving the Diffusability of Autoencoders](https://mlsanity.com/paper/2502.14831/) (2025)
- [Pythae: Unifying Generative Autoencoders in Python -- A Benchmarking Use Case](https://mlsanity.com/paper/2206.08309/) (2022)
- [Multi-Rate VAE: Train Once, Get the Full Rate-Distortion Curve](https://mlsanity.com/paper/2212.03905/) (2022)
- [The geometry of efficient codes: how rate-distortion trade-offs distort the latent representations of generative models](https://mlsanity.com/paper/2406.07269/) (2024)

## Abstract (arXiv)

Modern VAEs are rarely trained with the pointwise likelihood implied by the standard β-VAE objective. In practice, pointwise reconstruction is often combined with perceptual and adversarial losses, despite a lack of understanding of how this changes the latent dynamics of the model. We show that the choice of reconstruction loss reshapes the rate-distortion problem itself, altering both the information content and the geometry of the learned latent space in ways that may be invisible from reconstructions alone. First, we prove and verify empirically that augmenting pointwise reconstruction with neural terms, such as perceptual and adversarial objectives, reduces the amount of information stored in the latent representations. Second, we show that neural reconstruction losses systematically change the geometry of the latent space: they make representations more isotropic and distribute uncertainty more evenly across latent dimensions, producing different posterior variance profiles. These findings highlight how the rate-distortion tradeoff is not a comprehensive lens to understand the behavior of VAEs, and we propose a more mechanistic approach to investigate how the choice of a distortion metric reshapes the optimization problem.