Sparse-LaViDa: Sparse Multimodal Discrete Diffusion Language Models
This addresses the inference bottleneck for multimodal diffusion models, offering a practical speed improvement for users in image and text processing, though it is incremental as it builds on existing MDM frameworks.
The paper tackles the slow inference speed of Masked Discrete Diffusion Models (MDMs) by proposing Sparse-LaViDa, which dynamically truncates unnecessary masked tokens during sampling, achieving up to a 2x speedup across tasks like text-to-image generation and image editing while maintaining quality.
Masked Discrete Diffusion Models (MDMs) have achieved strong performance across a wide range of multimodal tasks, including image understanding, generation, and editing. However, their inference speed remains suboptimal due to the need to repeatedly process redundant masked tokens at every sampling step. In this work, we propose Sparse-LaViDa, a novel modeling framework that dynamically truncates unnecessary masked tokens at each inference step to accelerate MDM sampling. To preserve generation quality, we introduce specialized register tokens that serve as compact representations for the truncated tokens. Furthermore, to ensure consistency between training and inference, we design a specialized attention mask that faithfully matches the truncated sampling procedure during training. Built upon the state-of-the-art unified MDM LaViDa-O, Sparse-LaViDa achieves up to a 2x speedup across diverse tasks including text-to-image generation, image editing, and mathematical reasoning, while maintaining generation quality.