RegionReasoner: Region-Grounded Multi-Round Visual Reasoning
This work addresses the limitation of single-step or text-only reasoning in vision-language models for iterative visual understanding, though it appears incremental as it builds on existing reinforcement learning and grounding techniques.
The authors tackled the problem of multi-round visual reasoning by introducing a new benchmark and a reinforcement learning framework, RegionReasoner, which improved multi-round reasoning accuracy, spatial grounding precision, and global-local consistency in detection and segmentation tasks.
Large vision-language models have achieved remarkable progress in visual reasoning, yet most existing systems rely on single-step or text-only reasoning, limiting their ability to iteratively refine understanding across multiple visual contexts. To address this limitation, we introduce a new multi-round visual reasoning benchmark with training and test sets spanning both detection and segmentation tasks, enabling systematic evaluation under iterative reasoning scenarios. We further propose RegionReasoner, a reinforcement learning framework that enforces grounded reasoning by requiring each reasoning trace to explicitly cite the corresponding reference bounding boxes, while maintaining semantic coherence via a global-local consistency reward. This reward extracts key objects and nouns from both global scene captions and region-level captions, aligning them with the reasoning trace to ensure consistency across reasoning steps. RegionReasoner is optimized with structured rewards combining grounding fidelity and global-local semantic alignment. Experiments on detection and segmentation tasks show that RegionReasoner-7B, together with our newly introduced benchmark RegionDial-Bench, considerably improves multi-round reasoning accuracy, spatial grounding precision, and global-local consistency, establishing a strong baseline for this emerging research direction.