時間的GRPO:視覚言語行動強化学習における軌跡レベルの信用割当を超えて
Temporal GRPO: Beyond Trajectory-Level Credit in Vision-Language-Action Reinforcement Learning
軌跡全体に同じ報酬を割り当てる従来のGRPOの問題を解決するため、タスクを段階に分割し、同じ段階に到達した軌跡同士を比較して段階ごとの利点を計算する新しい手法を提案した。
詳しい要約
1. どんなもの?
2. 先行研究と比べてどこがすごい?
3. 技術・手法の肝は?
4. どうやって有効だと検証した?
5. 議論はある?
6. 次に読むべき論文は?
※ AIが要旨から生成した要約です。正確性は原文をご確認ください。
著者: Yao Zhou, Hang Gao, Fengge Wu, Changwen Zheng, Wenwen Qiang
分類: cs.RO
原文アブストラクト
Outcome-driven reinforcement learning offers a scalable way to post-train vision-language-action (VLA) policies from sparse task-success feedback. In common GRPO-based VLA post-training, one rollout-level advantage is applied to every action in the trajectory. A rollout that completes several valid stages but fails later can therefore penalize the actions that produced its earlier progress. We call this trajectory-level credit aliasing. Temporal GRPO addresses this problem by constructing detectable task stages, aligning each rollout with stage-specific action intervals, and comparing only rollouts that have entered the same stage. The resulting stage advantages are applied to their corresponding intervals in a single policy update. On RoboTwin 2.0, Temporal GRPO improves task success and sample efficiency, with consistent gains across task horizons. Controlled updates on LIBERO-Long preserve shared prerequisite stages and concentrate improvement at the first stage where rollout outcomes diverge.