Reusable Slotwise Mechanisms
This addresses the challenge of enabling agents to comprehend and reason about object interactions in novel scenarios, with incremental improvements in scene representation and dynamics modeling.
The paper tackles the problem of modeling object dynamics for improved robustness and generalization in agents by introducing Reusable Slotwise Mechanisms (RSM), a framework that uses modular, reusable mechanisms with Central Contextual Information to predict future states, achieving superior performance in future prediction and downstream tasks like Visual Question Answering and action planning compared to state-of-the-art methods.
Agents with the ability to comprehend and reason about the dynamics of objects would be expected to exhibit improved robustness and generalization in novel scenarios. However, achieving this capability necessitates not only an effective scene representation but also an understanding of the mechanisms governing interactions among object subsets. Recent studies have made significant progress in representing scenes using object slots. In this work, we introduce Reusable Slotwise Mechanisms, or RSM, a framework that models object dynamics by leveraging communication among slots along with a modular architecture capable of dynamically selecting reusable mechanisms for predicting the future states of each object slot. Crucially, RSM leverages the Central Contextual Information (CCI), enabling selected mechanisms to access the remaining slots through a bottleneck, effectively allowing for modeling of higher order and complex interactions that might require a sparse subset of objects. Experimental results demonstrate the superior performance of RSM compared to state-of-the-art methods across various future prediction and related downstream tasks, including Visual Question Answering and action planning. Furthermore, we showcase RSM's Out-of-Distribution generalization ability to handle scenes in intricate scenarios.