Self-contrastive alignment improves visual generation without extra guidance

SCALE: Improving Guidance-Free Visual Generation via Self-Contrastive Alignment for Likelihood Estimation

Computer Vision and Pattern Recognition

Summary

Generating high-quality images using AI often requires extra steps that slow down the process. The authors propose a new method that helps the AI learn directly from its own generated images by comparing them to real ones, improving image quality without added speed costs. Their approach, called SCALE, works in one step and outperforms existing methods that mimic the slower techniques. It also works well combined with other fine-tuning methods to make generated images even better.

What this means in practice

Authors

Jiwan Hur, Dong-Jae Lee, Dongyeun Lee, Gyojin Han, Junmo Kim

Abstract

Classifier-free guidance (CFG) is essential for improving sample quality in visual generation, but it incurs substantial sampling overhead. Existing guidance-free approaches halve this cost by approximating the CFG with a single model inference, but their performance is bounded by the CFG and inherits its limitations. In this paper, we propose self-contrastive guidance (SCG), which contrasts the model's prediction on real and self-generated samples to obtain a direct, model-based signal that pushes generation toward the true data manifold. Integrating SCG into the standard likelihood objective via reparameterization yields Self-Contrastive Alignment for Likelihood Estimation (SCALE), a guidance-free training framework that produces SCG-enhanced predictions in a single forward pass. Across class-conditional and text-to-image generation on diffusion and autoregressive models, SCALE consistently outperforms CFG-mimicking guidance-free baselines and improves further when combined with DPO, demonstrating that the self-contrastive signal yields gains complementary to both CFG and preference-based fine-tuning.