SNCE: Geometry-Aware Supervision for Scalable Discrete Image Generation

TL;DR

SNCE introduces a geometry-aware soft supervision method for large-codebook discrete image generation, improving training efficiency and image quality by capturing semantic structures in the embedding space.

Contribution

The paper proposes SNCE, a novel soft supervision training objective that enhances the optimization of large-codebook discrete image generators by leveraging geometric proximity.

Findings

SNCE accelerates convergence compared to standard methods.

SNCE improves image generation quality across multiple tasks.

SNCE effectively captures semantic geometric structures.

Abstract

Recent advancements in discrete image generation showed that scaling the VQ codebook size significantly improves reconstruction fidelity. However, training generative models with a large VQ codebook remains challenging, typically requiring larger model size and a longer training schedule. In this work, we propose Stochastic Neighbor Cross Entropy Minimization (SNCE), a novel training objective designed to address the optimization challenges of large-codebook discrete image generators. Instead of supervising the model with a hard one-hot target, SNCE constructs a soft categorical distribution over a set of neighboring tokens. The probability assigned to each token is proportional to the proximity between its code embedding and the ground-truth image embedding, encouraging the model to capture semantically meaningful geometric structure in the quantized embedding space. We conduct extensive experiments across class-conditional ImageNet-256 generation, large-scale text-to-image synthesis, and image editing tasks. Results show that SNCE significantly improves convergence speed and overall generation quality compared to standard cross-entropy objectives.