Omni-R1: Towards the Unified Generative Paradigm for Multimodal Reasoning

TL;DR

This work reframes multimodal reasoning as unified generative processing, introducing Omni-R1, a two-stage SFT+RL model with perception-aligned supervision, and Omni-R1-Zero, which bootstraps step-wise visual reasoning from text-only data. The approach targets diverse Uni-Tasks through a common generative framework that produces intermediate functional images corresponding to reasoning steps. Empirical results on Omni-Bench and broader multimodal benchmarks show consistent gains over strong baselines, with Omni-R1 excelling when reliable intermediate visual evidence is crucial and Omni-R1-Zero approaching or surpassing Omni-R1 under zero-shot supervision. The findings suggest a promising direction toward scalable, generalizable generative multimodal reasoning that reduces annotation costs while maintaining interpretable intermediate representations.

Abstract

Multimodal Large Language Models (MLLMs) are making significant progress in multimodal reasoning. Early approaches focus on pure text-based reasoning. More recent studies have incorporated multimodal information into the reasoning steps; however, they often follow a single task-specific reasoning pattern, which limits their generalizability across various multimodal tasks. In fact, there are numerous multimodal tasks requiring diverse reasoning skills, such as zooming in on a specific region or marking an object within an image. To address this, we propose unified generative multimodal reasoning, which unifies diverse multimodal reasoning skills by generating intermediate images during the reasoning process. We instantiate this paradigm with Omni-R1, a two-stage SFT+RL framework featuring perception alignment loss and perception reward, thereby enabling functional image generation. Additionally, we introduce Omni-R1-Zero, which eliminates the need for multimodal annotations by bootstrapping step-wise visualizations from text-only reasoning data. Empirical results show that Omni-R1 achieves unified generative reasoning across a wide range of multimodal tasks, and Omni-R1-Zero can match or even surpass Omni-R1 on average, suggesting a promising direction for generative multimodal reasoning.

Figures (9)

• Figure 1: An example illustrating the necessity of incorporating visual information into the intermediate reasoning steps for multimodal tasks.
• Figure 2: The illustration of various multimodal tasks and the corresponding diverse multimodal reasoning skills, with multimodal tasks shown in the top row and the required reasoning skills summarized in the bottom row.
• Figure 3: Training pipeline for Omni-R1 and Omni-R1-Zero.(Top) Data initialization uses either limited human annotations (orange) or synthetic bootstrapped trajectories (blue). (Middle) Stage 1 (PeSFT) performs supervised fine-tuning with joint cross-entropy and perception losses. (Bottom) Stage 2 (PeRPO) refines the policy using unified tasks without multimodal annotation (left) and a composite reward (Accuracy, Format, Perception) for final alignment.
• Figure 4: The t-SNE visualization of generated images from Omni-R1 and Omni-R1-Zero, respectively. Filled markers indicate correct predictions and empty markers indicate incorrect predictions.
• Figure 5: Case studies of Omni-R1's multimodal reasoning skills, including (a) Grounding (Zoom-In) for attribute recognition, (b) Grounding (BBOX) for target localization, and (c) Grounding (BBOX) for parameter localization. (d) Marking for graph connectivity verification and (e) Visual prediction for robotic manipulation sequences.