VLA-Trace: 表現と行動のトレーシングによる視覚-言語-行動モデルの診断
VLA-Trace: Diagnosing Vision-Language-Action Models through Representation and Behavior Tracing
VLAモデルがマルチモーダル知識を身体制御に変換する過程を、表現ダイナミクスから因果帰属、行動発現まで段階的に診断するフレームワークを提案し、π0.5とOpenVLAの比較分析からモデル間の適応戦略や制御経路の違いを明らかにした。
著者: Haoyuan Shi, Xiancong Ren, Yingji Zhang, Qinfan Zhang, Jiayu Hu, Haozhe Shan, Han Dong, Jinpeng Lu, Yinda Chen, Yi Zhang, Yong Dai, Xiaozhu Ju
分類: cs.AI
原文アブストラクト
Understanding how Vision-Language-Action (VLA) models transform multimodal knowledge into embodied control remains an open challenge. We present VLA-Trace, a progressive diagnostic framework that analyzes VLA models through a unified evidence chain from representation dynamics to causal control attribution and behavioral manifestation. It specifically combines cross-modal and checkpoint-drift centered kernel alignment (CKA) to trace representation evolution, attention knockout interventions to identify modality-specific control pathways, and rollout-level behavioral probes to examine grounding, shortcut dependence, and semantic following. Experiments on $π_{0.5}$ and OpenVLA reveal three key findings. First, the two models exhibit distinct modality-specific adaptation dynamics during VLA finetuning. Second, they rely on different multimodal routing strategies and layer-wise dependencies during action decoding. Third, although VLA policies excel at visually grounded trajectory generation, they remain limited in fine-grained semantic following. These findings highlight future directions for representation-preserving adaptation, causal VLA circuits, and compositional semantic control.