Newest papers

PICA-WAM: Visual Perturbation Invariance via Cross-Appearance Self-Alignment in World Action Models

Yude Zou, Yifei Yao, Guanbo Jiao, Jianming Ma, Guanjie Zheng, Xing Gao · 2026 · arXiv (Cornell University)

DOI: 10.48550/arxiv.2610.26813 · cited by 0 works (OpenAlex) · open-access PDF

Spanish

Los modelos de acción mundial (WAMs) predicen conjuntamente el vídeo futuro y generan acciones, pero siguen siendo sensibles a las perturbaciones visuales irrelevantes a la tarea, es decir, cambios en la apariencia visual que no deben afectar al comportamiento del robot. Identificamos una brecha en el entrenamiento convencional de WAM: la predicción de vídeo alienta las representaciones para preservar la información de apariencia pre-entrenada, mientras que la generación de acción no impone explícitamente la invariancia a la apariencia irrelevante a la tarea. La ampliación de datos expone a las WAMs a tales perturbaciones, pero no requiere observaciones perturbadas y nominales para producir representaciones de acción consistentes, y el alineamiento con las características sensoriales externas no garantiza la invariancia deseada

Machine translation (Meta M2M100); the English original follows.

Abstract

World Action Models (WAMs) jointly predict future video and generate actions, yet remain sensitive to task-irrelevant visual perturbations, i.e., changes in visual appearance that should not affect robot behavior. We identify a gap in conventional WAM training: video prediction encourages representations to preserve appearance information, while action generation does not explicitly enforce invariance to task-irrelevant appearance. Data augmentation exposes WAMs to such perturbations but does not require perturbed and nominal observations to produce consistent action representations, and alignment with external pretrained features does not guarantee the desired invariance. To address this gap, we propose PICA-WAM, a self-supervised training framework that casts perturbation invariance as an auxiliary cross-appearance self-alignment task for WAMs. An exponential-moving-average (EMA) teacher processes the nominal observation to provide a deeper-layer target, while the online WAM processes a perturbed observation and aligns its shallower-layer action representation to this target through a lightweight projection head. Across four strong WAMs evaluated on RoboTwin 2.0, LIBERO-Plus, EBench, and a real robot, PICA-WAM consistently improves robustness under visual perturbations, preserves nominal performance in simulation, and raises the real robot success rate. PICA-WAM further reduces the sensitivity of video features to task-irrelevant appearance and its propagation into actions, as shown by paired layer-wise diagnostics. Code: https://github.com/YudeZou/PICA-WAM-release

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Topics: Robustness (evolution), Artificial intelligence, Perturbation (astronomy), Computer science, Computer vision, Robot

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange