On Memory: A comparison of memory mechanisms in world models
Fuente:
arXiv
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866908697608847360 |
|---|---|
| author | Laird, Eli J. Clark, Corey |
| author_facet | Laird, Eli J. Clark, Corey |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_06983 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | On Memory: A comparison of memory mechanisms in world models Laird, Eli J. Clark, Corey Artificial Intelligence Machine Learning 68T45 (Primary) 68T07 (Secondary) I.2.10 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. |
| title | On Memory: A comparison of memory mechanisms in world models |
| topic | Artificial Intelligence Machine Learning 68T45 (Primary) 68T07 (Secondary) I.2.10 |
| url | https://arxiv.org/abs/2512.06983 |