Dr. Seg: Revisiting GRPO Training for Visual Large Language Models through Perception-Oriented Design

TL;DR

This paper introduces Dr. Seg, a perception-oriented training framework for Visual Large Language Models that addresses limitations of language reasoning paradigms, enhancing performance in visual perception tasks without architectural changes.

Contribution

The paper identifies key differences between reasoning and perception training paradigms and proposes a novel, plug-and-play GRPO-based framework called Dr. Seg for improved visual perception.

Findings

Dr. Seg improves VLLM performance in complex visual tasks.

It maintains strong generalization across diverse visual scenarios.

No architectural modifications are required for integration.

Abstract

Following the success of Group Relative Policy Optimization (GRPO) in foundation LLMs, an increasing number of works have sought to adapt GRPO to Visual Large Language Models (VLLMs) for visual perception tasks (e.g., detection and segmentation). However, much of this line of research rests on a long-standing yet unexamined assumption: training paradigms developed for language reasoning can be transferred seamlessly to visual perception. Our experiments show that this assumption is not valid, revealing intrinsic differences between reasoning-oriented and perception-oriented settings. Using reasoning segmentation as a representative case, we surface two overlooked factors: (i) the need for a broader output space, and (ii) the importance of fine-grained, stable rewards. Building on these observations, we propose Dr.~Seg, a simple, plug-and-play GRPO-based framework consisting of a Look-to-Confirm mechanism and a Distribution-Ranked Reward module, requiring no architectural modifications and integrating seamlessly with existing GRPO-based VLLMs. Extensive experiments demonstrate that Dr.~Seg improves performance in complex visual scenarios while maintaining strong generalization. Code, models, and datasets are available at https://github.com/eVI-group-SCU/Dr-Seg.