PhysTrace：動画生成モデルの物理的整合性を時間的に解釈評価するフレームワーク
PhysTrace: Interpretable Temporal Evaluation of Physical Consistency in Video Generative Models
MLLMが生成した履歴条件付きの物理的期待と観測動画を照合し、物理法則違反の箇所を時間的に特定・説明する学習不要の評価手法を提案。IntPhys2に違反区間を付与したIntPhys2-Timingも導入した。
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: Taekoan Yoo, Wonkyung Jung, Kyunghun Kim, Kyeongbo Kong
分類: cs.AI, cs.RO
原文アブストラクト
World action models increasingly rely on pretrained video models assumed to capture physical dynamics for prediction, planning, and action generation. Existing evaluations assess plausibility, probe representations, or measure predictive surprise, but rarely identify which history-conditioned physical expectations conflict with observations. We introduce PhysTrace, a training-free framework aligning MLLM-generated, history-conditioned expectations with observed video windows to temporally localize physical violations and provide interpretable descriptions without task-specific probes or classifiers. We also introduce IntPhys2-Timing, adding violation intervals to IntPhys2. PhysTrace achieves a Hit Rate of up to 70.7% across physical principles on IntPhys2-Timing and outperforms the compared methods in overall possible-impossible discrimination accuracy on IntPhys2. Applied to reconstructed videos along the trajectories of pretrained diffusion/flow-based video models, PhysTrace reveals attribute- and noise-dependent patterns in physical violation localization, enabling interpretable physical consistency analysis across timesteps.