JEPA世界モデルによる価値誘導型行動計画
Value-guided action planning with JEPA world models
JEPA世界モデルの表現空間を整形し、目標条件付き価値関数を状態埋め込み間の距離で近似することで、行動計画の性能を向上させる手法を提案した。
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: Matthieu Destrade, Oumayma Bounou, Quentin Le Lidec, Jean Ponce, Yann LeCun
分類: cs.LG, cs.AI, cs.RO
原文アブストラクト
Building deep learning models that can reason about their environment requires capturing its underlying dynamics. Joint-Embedded Predictive Architectures (JEPA) provide a promising framework to model such dynamics by learning representations and predictors through a self-supervised prediction objective. However, their ability to support effective action planning remains limited. We propose an approach to enhance planning with JEPA world models by shaping their representation space so that the negative goal-conditioned value function for a reaching cost in a given environment is approximated by a distance (or quasi-distance) between state embeddings. We introduce a practical method to enforce this constraint during training and show that it leads to significantly improved planning performance compared to standard JEPA models on simple control tasks.
