LaViDa-R1: Advancing Reasoning for Unified Multimodal Diffusion Language Models

TL;DR

LaViDa-R1 is a versatile multimodal diffusion language model that unifies various reasoning tasks through a novel training framework, achieving strong performance across diverse multimodal applications.

Contribution

It introduces a unified post-training framework combining supervised fine-tuning and multi-task reinforcement learning for multimodal reasoning.

Findings

Strong performance on visual math reasoning

Effective in reason-intensive grounding tasks

Excels in image editing applications

Abstract

Diffusion language models (dLLMs) recently emerged as a promising alternative to auto-regressive LLMs. The latest works further extended it to multimodal understanding and generation tasks. In this work, we propose LaViDa-R1, a multimodal, general-purpose reasoning dLLM. Unlike existing works that build reasoning dLLMs through task-specific reinforcement learning, LaViDa-R1 incorporates diverse multimodal understanding and generation tasks in a unified manner. In particular, LaViDa-R1 is built with a novel unified post-training framework that seamlessly integrates supervised finetuning (SFT) and multi-task reinforcement learning (RL). It employs several novel training techniques, including answer-forcing, tree search, and complementary likelihood estimation, to enhance effectiveness and scalability. Extensive experiments demonstrate LaViDa-R1's strong performance on a wide range of multimodal tasks, including visual math reasoning, reason-intensive grounding, and image editing.