適応手法の選び方：Vision-Language-Actionモデルのためのデータ別ガイド
Post-Training, Finetuning, and Steering:A Field Guide for Curating Training Data
VLAモデルの適応手法を、オン/オフポリシーデータやオフライン/オンライン適応などのデータ条件で比較し、状況に応じた最適な手法選択の指針を示した。
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: Maria Attarian, Yifan Ruan, Claas Voelcker, Chenyang Cao, Amy Zhang, Peter Stone, Nicholas Rhinehart, Igor Gilitschenski
分類: cs.RO
原文アブストラクト
While Vision-Language-Action models have enabled impressive cross-task transfer in zero-shot settings, \emph{adaptation} methods are often necessary to achieve consistent performance on novel tasks. However, even though a wealth of methods for adaptation have been proposed, it is unclear how to choose the best method based on available trajectories. We close this gap by studying the impact of the adaptation method under a variety of common data scenarios: on- and off-policy data, offline vs online adaptation, and whether the data contains optimal or suboptimal trajectories. Our evaluation spans simulated tasks in the LIBERO suite and real tasks on a Franka platform. In the offline data regime, critic-based steering approaches outperform supervised finetuning when failure data is available. However, offline RL techniques for training critic functions fail to outperform simpler Monte-Carlo-based approaches. Finally, in the online setting, critic-based steering is significantly more sample efficient compared to policy-gradient-based finetuning methods, but has a lower performance ceiling than full finetuning.
