On Memory: A comparison of memory mechanisms in world models
By: Eli J. Laird, Corey Clark
Potential Business Impact:
Helps AI remember longer to plan better.
World models enable agents to plan within imagined environments by predicting future states conditioned on past observations and actions. However, their ability to plan over long horizons is limited by the effective memory span of the backbone architecture. This limitation leads to perceptual drift in long rollouts, hindering the model's capacity to perform loop closures within imagined trajectories. In this work, we investigate the effective memory span of transformer-based world models through an analysis of several memory augmentation mechanisms. We introduce a taxonomy that distinguishes between memory encoding and memory injection mechanisms, motivating their roles in extending the world model's memory through the lens of residual stream dynamics. Using a state recall evaluation task, we measure the memory recall of each mechanism and analyze its respective trade-offs. Our findings show that memory mechanisms improve the effective memory span in vision transformers and provide a path to completing loop closures within a world model's imagination.
Similar Papers
Memory-Augmented Transformers: A Systematic Review from Neuroscience Principles to Enhanced Model Architectures
Machine Learning (CS)
Helps computers remember more, like humans.
Memory-Augmented Transformers: A Systematic Review from Neuroscience Principles to Technical Solutions
Machine Learning (CS)
Computers remember more, learn longer, and think better.
Video World Models with Long-term Spatial Memory
CV and Pattern Recognition
Keeps computer-made videos consistent over time.