回復から低下へ：アクション事後学習がVLMの後層深度デコード可能性を減少させる仕組み
From Recovery to Drop-off: How Action Post-training Reduces a VLM's Late-Layer Depth Decodability
VLA構築のためのアクション事後学習が、VLMの空間理解（特に深度知覚）をどのように劣化させるかを、層ごとのプローブで分析した論文。後層のMLP干渉が主因であることを特定した。
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: Alexander Hackett, Arnaud Denis-Remillard, Axel Cassou
分類: cs.CV, cs.AI, cs.LG
原文アブストラクト
How much of a vision-language model's (VLM) spatial understanding remains after the action post-training process of building a vision-language-action model (VLA)? We probe depth perception, a primitive of spatiogeometric understanding, from every decoder layer of a weight-matched open-source base VLM/VLA pair: Molmo2-ER and MolmoAct2-LIBERO. First, the VLA decodes depth worse at every layer, a persistent gap we call the floor. Second, the degradation is not uniform: while the base VLM's depth decodability improves through its final layers, the VLA's collapses, an additional late-layer drop we call the cliff. We causally localize the cliff to late-layer MLP interference: ablating the late-layer MLP writes recovers the majority of the terminal decodability cliff, while matched attention ablations and the same intervention in the weight-matched base VLM produce no comparable recovery. A module-level decomposition explains this dissociation: the base VLM carries depth most accessibly in accumulated MLP writes, whereas action post-training collapses depth decodability in the late accumulated writes.