跳到正文
研踪PaperTrail
← 研踪 PaperTrail

arXiv:2605.23500 · v3

GRTO: Bootstrapped Group Relative Tool Optimization for Referring Segmentation

首次提交 2026-05-22 19:04
版本更新 2026-11-18 14:54
雷达观测 2026-11-19 19:08
arXiv ↗PDF ↗

摘要

Referring segmentation requires reasoning to connect language with visual evidence and identify the intended target in challenging scenarios. This makes GRPO a natural choice for training VLMs together with a segmentation tool. However, existing approaches optimize only the VLM, assuming that the tool is perfect and leaving it untouched. This assumption is particularly limiting when the tool faces domain shift, as segmentation errors can undermine even correct reasoning. We therefore argue that the segmentation tool should be optimized together with the reasoning policy. However, this is challenging because policy optimization and differentiable tool learning rely on fundamentally different objectives. In this work, we introduce group relative tool optimization (GRTO), a mathematically grounded two-stage framework for jointly optimizing a policy and its differentiable tool. The key idea is to first align the tool with the policy expected after joint training, rather than the initial policy. The prealigned tool then leads to a more robust reward signal and superior policy convergence in subsequent policy training. In the first stage, GRTO bootstraps the tool toward the theoretically optimal policy, capturing readily attainable gains without costly on-policy rollouts. This alignment bootstraps the second stage, where GRTO extends GRPO to jointly refine the policy and tool. Across three challenging referring segmentation benchmarks, GRTO achieves substantial improvements over plain GRPO and alternative tool training procedures, 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.

AI 论文解读

基于摘要 · 非全文审阅

此版本尚未生成 AI 解读。解读由后台任务生成并按论文版本保存,登录后共享已有结果。

版本差异

本站首次收录为 v3,暂无更早版本可对照。

观测到的版本

相邻记录可能跳过中间版本;首次发现 v2 及以上也会标为版本更新。

v3 · 2026-11-18 14:54 · 最新观测版本

GRTO: Bootstrapped Group Relative Tool Optimization for Referring Segmentation

Referring segmentation requires reasoning to connect language with visual evidence and identify the intended target in challenging scenarios. This makes GRPO a natural choice for training VLMs together with a segmentation tool. However, existing approaches optimize only the VLM, assuming that the tool is perfect and leaving it untouched. This assumption is particularly limiting when the tool faces domain shift, as segmentation errors can undermine even correct reasoning. We therefore argue that the segmentation tool should be optimized together with the reasoning policy. However, this is challenging because policy optimization and differentiable tool learning rely on fundamentally different objectives. In this work, we introduce group relative tool optimization (GRTO), a mathematically grounded two-stage framework for jointly optimizing a policy and its differentiable tool. The key idea is to first align the tool with the policy expected after joint training, rather than the initial policy. The prealigned tool then leads to a more robust reward signal and superior policy convergence in subsequent policy training. In the first stage, GRTO bootstraps the tool toward the theoretically optimal policy, capturing readily attainable gains without costly on-policy rollouts. This alignment bootstraps the second stage, where GRTO extends GRPO to jointly refine the policy and tool. Across three challenging referring segmentation benchmarks, GRTO achieves substantial improvements over plain GRPO and alternative tool training procedures, 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.

查看此版本解读 →查看此版本原文 ↗