Safe Autoregressive Image Generation with Iterative Self-Improving Codebooks

2026-06-25

AI Summary

This work addresses the vulnerability of autoregressive multimodal image generation models to producing unsafe content and their lack of intrinsic safety mechanisms that operate without human annotation. To this end, the authors propose an iterative self-optimizing codebook approach that leverages the self-discriminative capability of a unified multimodal model to identify harmful images and construct paired safe and unsafe text-image samples. By alternately pruning harmful regions of the codebook space and adaptively fine-tuning the benign regions, the method enables continuous safety-aware updates to the codebook. Notably, this is the first approach to achieve endogenous safety optimization in autoregressive image generation models without relying on external feedback or manual labeling, significantly reducing the generation of harmful content while preserving high-fidelity image synthesis capabilities.

Abstract

Unlike diffusion-based models that operate in continuous latent spaces, autoregressive unified multimodal models produce images by sequentially predicting discretized visual tokens. These tokens are derived from a codebook that maps embeddings to quantized visual patterns. The language-like architecture enables unified multimodal models to effectively capture text conditional information for generation, making them promising for text-to-image tasks. This also raises an interesting question: how safe are the images generated in such an autoregressive way? In this work, we propose iterative self-improving codebooks for safe autoregressive generation. We leverage the understanding and judgment capabilities of the unified multimodal model itself to identify unsafe generated images without human annotation. Subsequently, the inherent representations in the codebook are fixed to eliminate harmful mappings. Our method comprises two steps: first, we use the unified model to identify unsafe generations and construct corresponding harmful and safe image-text pairs. These pairs are used to construct the Harmful Space and guide updates to the codebook, thereby eliminating harmful outputs. Second, we perform adaptive fine-tuning on the codebook within the harmless space using safe image-text pairs to ensure the quality of generated images. These two steps are repeated until no further improvement is observed, producing a safety-enhanced model codebook. Without additional external feedback, the safety of models is improved iteratively.

Problem

• autoregressive image generation
• safety
• codebook
• multimodal models
• harmful content

Innovation

• iterative self-improving codebooks
• safe autoregressive generation
• unified multimodal models
• harmful space
• codebook fine-tuning