A Minimal Task Reveals Emergent Path Integration and Object-Location Binding in a Predictive Sequence Model
Linda Ariel Ventura, Victoria Bosch, Tim C Kietzmann, Sushrut Thorat · 2026
Read A Minimal Task Reveals Emergent Path Integration and Object-Location Binding in a Predictive Sequence Model by Linda Ariel Ventura, Victoria Bosch, Tim C Kietzmann, Sushrut Thorat (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-03490.txt.
Description
How does a brain keep track of where things are while the eyes jump from place to place? Linda Ariel Ventura and 3 co-authors explore that question with a small neural network in a 2026 paper. The network reads tokens from two-dimensional scenes one at a time. It must predict the next token, given the current one and a saccade-like jump.
On fresh scenes, predictions improve as the sequence proceeds, showing in-context learning. Decoding the network's activity reveals path integration, a running tally of position, and dynamic binding of each token's identity to its location. Interventions show new bindings can form late in a sequence, even for unusual combinations. The results offer a concrete mechanism for sequential world modeling.
Catalog record
| CREATOR | Linda Ariel Ventura, Victoria Bosch, Tim C Kietzmann, Sushrut Thorat |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 25 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-03490