Pith. sign in

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

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it
abstract

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.

fields

cs.CV 1

years

2026 1

verdicts

CONDITIONAL 1

representative citing papers

More with Less: a Large Scale Remote Sensing VLM with a Simple Recipe

cs.CV · 2026-07-17 · conditional · novelty 5.0

An unmodified general-purpose VLM trained with multi-task RL and a SAM3 tool reaches top results on most remote sensing zero-shot benchmarks, with gains the paper attributes to training-data diversity rather than architecture.

citing papers explorer

Showing 1 of 1 citing paper.

  • More with Less: a Large Scale Remote Sensing VLM with a Simple Recipe cs.CV · 2026-07-17 · conditional · none · ref 26 · internal anchor

    An unmodified general-purpose VLM trained with multi-task RL and a SAM3 tool reaches top results on most remote sensing zero-shot benchmarks, with gains the paper attributes to training-data diversity rather than architecture.