DiverseAR: Boosting Diversity in Bitwise Autoregressive Image Generation
By: Ying Yang , Zhengyao Lv , Tianlin Pan and more
Potential Business Impact:
Makes AI create more varied and realistic pictures.
In this paper, we investigate the underexplored challenge of sample diversity in autoregressive (AR) generative models with bitwise visual tokenizers. We first analyze the factors that limit diversity in bitwise AR models and identify two key issues: (1) the binary classification nature of bitwise modeling, which restricts the prediction space, and (2) the overly sharp logits distribution, which causes sampling collapse and reduces diversity. Building on these insights, we propose DiverseAR, a principled and effective method that enhances image diversity without sacrificing visual quality. Specifically, we introduce an adaptive logits distribution scaling mechanism that dynamically adjusts the sharpness of the binary output distribution during sampling, resulting in smoother predictions and greater diversity. To mitigate potential fidelity loss caused by distribution smoothing, we further develop an energy-based generation path search algorithm that avoids sampling low-confidence tokens, thereby preserving high visual quality. Extensive experiments demonstrate that DiverseAR substantially improves sample diversity in bitwise autoregressive image generation.
Similar Papers
Diversity Has Always Been There in Your Visual Autoregressive Models
CV and Pattern Recognition
Makes AI create more varied and interesting pictures.
Towards Better & Faster Autoregressive Image Generation: From the Perspective of Entropy
CV and Pattern Recognition
Makes AI draw better pictures, faster.
DiverseVAR: Balancing Diversity and Quality of Next-Scale Visual Autoregressive Models
CV and Pattern Recognition
Makes AI create more different pictures from the same words.