Mixture-of-Experts VLAにおける創発的な構成スキル
Emergent Compositional Skills in Mixture-of-Experts VLAs
専門家デモからエンドツーエンドでロボットポリシーを学習する際、Mixture-of-Expertsアーキテクチャがタスクを再利用可能なプリミティブに自動分解できることを示した論文。
著者: Shlok Shah, Rhiaan Jhaveri, Tharun Kumar Tiruppali Kalidoss, Chirayu Nimonkar, Ishaan Javali, Dhruv Shah
分類: cs.RO, cs.AI, cs.LG
原文アブストラクト
We consider the problem of learning compositional robot policies end-to-end from expert demonstrations, without any pre-specified notion of task decomposition or hierarchy. We ask whether a VLA trained with a simplified Mixture-of-Experts (MoE) action head can emergently learn to decompose tasks into reusable, interpretable primitives. We find that learned experts are heavily reused across tasks and consistently correspond to qualitatively distinct low-level behaviors, suggesting that the router implicitly learns to perform high-level sequencing while experts serve as compositional primitives. Our MoE matches the task performance of a monolithic baseline while demonstrating meaningful expert specialization, a step toward modular, interpretable robot policies that emerge from data alone.