LaViDa: A Large Diffusion Language Model for Multimodal Understanding

TL;DR

LaViDa introduces a diffusion-based vision-language model that offers faster inference, controllable generation, and bidirectional reasoning, outperforming autoregressive models on key multimodal benchmarks.

Contribution

This paper presents LaViDa, a novel multimodal VLM built on diffusion models, with techniques for effective training, inference efficiency, and improved multimodal understanding.

Findings

Outperforms AR VLMs on MMMU benchmark

Surpasses Open-LLaVa-Next-8B on COCO captioning by +4.1 CIDEr

Achieves +59% improvement on Constrained Poem Completion

Abstract

Modern Vision-Language Models (VLMs) can solve a wide range of tasks requiring visual reasoning. In real-world scenarios, desirable properties for VLMs include fast inference and controllable generation (e.g., constraining outputs to adhere to a desired format). However, existing autoregressive (AR) VLMs like LLaVA struggle in these aspects. Discrete diffusion models (DMs) offer a promising alternative, enabling parallel decoding for faster inference and bidirectional context for controllable generation through text-infilling. While effective in language-only settings, DMs' potential for multimodal tasks is underexplored. We introduce LaViDa, a family of VLMs built on DMs. We build LaViDa by equipping DMs with a vision encoder and jointly fine-tune the combined parts for multimodal instruction following. To address challenges encountered, LaViDa incorporates novel techniques such as complementary masking for effective training, prefix KV cache for efficient inference, and timestep shifting for high-quality sampling. Experiments show that LaViDa achieves competitive or superior performance to AR VLMs on multi-modal benchmarks such as MMMU, while offering unique advantages of DMs, including flexible speed-quality tradeoff, controllability, and bidirectional reasoning. On COCO captioning, LaViDa surpasses Open-LLaVa-Next-8B by +4.1 CIDEr with 1.92x speedup. On bidirectional tasks, it achieves +59% improvement on Constrained Poem Completion. These results demonstrate LaViDa as a strong alternative to AR VLMs. Code and models will be released in the camera-ready version.