スパースオートエンコーダがVLAモデルの解釈可能で操作可能な特徴を明らかにする
Sparse Autoencoders Reveal Interpretable and Steerable Features in VLA Models
VLAモデルの隠れ層にスパースオートエンコーダを適用し、動作プリミティブや意味概念に対応する解釈可能な特徴を発見。特徴の増幅・除去による操作実験で、汎化と制御可能性を実証した。
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: Aiden Swann, Lachlain McGranahan, Hugo Buurmeijer, Monroe Kennedy, Mac Schwager
分類: cs.RO
原文アブストラクト
Vision-Language-Action (VLA) models have emerged as a promising approach for general-purpose robot manipulation. However, little research has mechanistically explored when and why they generalize across objects, scenes, and instructions. To probe internal representations, we train Sparse Autoencoders (SAEs) on the VLA's hidden-layer activations. SAEs learn sparse dictionaries over model activations, often revealing features that correspond to interpretable directions in the model's representation space. We identify SAE features corresponding to motion primitives and semantic concepts, including features that are general across episodes and causally steerable. We propose a metric to categorize features as general transferable primitives or episode-specific memorizations, offering a promising glimpse towards VLA generalization. We validate these findings through steering experiments on both the LIBERO simulation benchmark and on real-world DROID hardware. We find that amplifying general and semantic features induces behaviors consistent with their meanings, whereas ablating them destroys model performance. Furthermore, we demonstrate steering as a way to control behavior in unpromptable directions. Together, these results provide mechanistic evidence that VLAs can learn reusable internal features linking perception, language, and action across tasks and scenes. Our project page is located at https://drvla.github.io
関連論文
- 視覚・言語・行動モデルの潜在クラスタ解析VLA/解釈可能性
- VLA表現からのタスク進捗の解読VLA/解釈可能性
- 具現化解釈可能性：視覚-言語-行動モデルにおける因果理解と汎化の関連付けVLA/解釈可能性
- 視覚-言語-行動モデルにおける特徴の観測と制御VLA/解釈可能性
