Path Integration and Object-Location Binding Emerge in an Action-Conditioned Predictive Sequence Network

TL;DR

This study demonstrates how a recurrent neural network can develop structured, flexible representations of objects and their relations through sequence prediction, supporting in-context learning and dynamic binding.

Contribution

The paper provides a mechanistic account of how structured, flexible object representations and path integration emerge in an action-conditioned predictive sequence network.

Findings

Prediction accuracy improves with sequence length on novel scenes.

Decoding reveals emergence of path integration and object-position binding.

New object bindings can be learned late and out-of-distribution bindings can be acquired.

Abstract

Adaptive cognition requires structured internal models of objects and their relations. Predictive neural networks are often proposed to learn such world models, but how these are instantiated and how they support prediction remain unclear. We investigate this in a minimal in-silico setting. A recurrent neural network samples tokens sequentially from 2D continuous token scenes and is trained to predict the upcoming token from the current input and a saccade-like displacement. On novel scenes, prediction accuracy improves across the sequence, indicating in-context learning. Decoding analyses reveal path integration and dynamic binding of token identity to position. Interventional analyses show that new bindings can be learned late in sequence and that out-of-distribution bindings can be learned as well. Together, these findings show how structured representations relying on flexible binding emerge to support prediction, offering a mechanistic account of sequential world modeling relevant to cognitive science.