Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

B-GRTO: Bootstrapped Group Relative Tool Optimization for Referring Segmentation

About

Segmentation is a fundamental task in computer vision, underpinning pixel-level scene understanding and serving as a cornerstone for applications ranging from autonomous perception to medical image analysis. For complex referring segmentation, recent methods pair large vision-language models with segmentation decoders: the former analyzes the image and prompt, while the latter predicts the target mask. Although reinforcement learning improves reasoning-intensive vision-language systems, trainable tools such as segmentation decoders are typically optimized separately with differentiable objectives, and the principled integration of such objectives into reinforcement learning remains underexplored. Thus, we introduce group relative tool optimization (GRTO), a mathematically grounded framework for jointly optimizing a policy with differentiable tool use. GRTO reuses group relative policy optimization (GRPO) rollouts to optimize the auxiliary tool objective, letting decoder gradients complement policy rewards. Further, we derive Bootstrapped-GRTO (B-GRTO), a pre-training method that cheaply bootstraps the tool, leading to faster convergence and superior performance. Across three challenging referring segmentation settings, B-GRTO results in substantial improvements over plain GRPO, matching or surpassing domain-specific state-of-the-art methods. This demonstrates the value of unifying reinforcement learning with differentiable auxiliary objectives for reasoning-intensive segmentation.

Mario Markov, Stefan Maria Ailuro, Mohammad Mahdi, Luc Van Gool, Danda Pani Paudel• 2026

Related benchmarks

TaskDatasetResultRank
Camouflaged Object DetectionCOD10K (test)
S-measure (S_alpha)0.907
306
Reasoning SegmentationReasonSeg (test)
gIoU68.9
236
Camouflaged Object DetectionNC4K (test)--
89
Reasoning SegmentationEarthReason (val)
gIoU76.35
47
Reasoning SegmentationEarthReason (test)
gIoU76.86
28
Camouflage DetectionCAMO (test)
S Score87.6
13
Reasoning SegmentationReasonSeg-R (test)
gIoU72
12
Reasoning SegmentationReasonSeg-X (val)
gIoU52.1
12
Reasoning SegmentationReasonSeg-X (test)
gIoU48.2
12
Showing 9 of 9 rows

Other info

Follow for update