Physical Reasoning Using Dynamics-Aware Models
This addresses a problem in AI for physical reasoning by improving model performance on benchmark tasks, though it is incremental as it builds on existing value learning approaches.
The study tackled the limitation of learning object dynamics solely from final reward values in physical reasoning tasks by augmenting rewards with self-supervised signals about dynamics, leading to substantial performance improvements and establishing a new state-of-the-art on the PHYRE benchmark.
A common approach to solving physical reasoning tasks is to train a value learner on example tasks. A limitation of such an approach is that it requires learning about object dynamics solely from reward values assigned to the final state of a rollout of the environment. This study aims to address this limitation by augmenting the reward value with self-supervised signals about object dynamics. Specifically, we train the model to characterize the similarity of two environment rollouts, jointly with predicting the outcome of the reasoning task. This similarity can be defined as a distance measure between the trajectory of objects in the two rollouts, or learned directly from pixels using a contrastive formulation. Empirically, we find that this approach leads to substantial performance improvements on the PHYRE benchmark for physical reasoning (Bakhtin et al., 2019), establishing a new state-of-the-art.