Guardado en:
Detalles Bibliográficos
Autores principales: Tang, Jinzhou, Liu, Sidi, Xiu, Waikit, Chen, Weixing, Wang, Keze
Formato: Preprint
Publicado: 2026
Materias:
Acceso en línea:https://arxiv.org/abs/2605.16899
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866910226099208192
author Tang, Jinzhou
Liu, Sidi
Xiu, Waikit
Chen, Weixing
Wang, Keze
author_facet Tang, Jinzhou
Liu, Sidi
Xiu, Waikit
Chen, Weixing
Wang, Keze
contents A fundamental challenge in embodied AI is verifying if agents build internal models of spatial structure or merely learn to mimic task-specific expert trajectories. This is critical as foundational approaches rooted in action-centric tasks (e.g., VLN) and reasoning-centric tasks (e.g., EQA) often share a common limitation: they lack a learning signal that forces them to encode fine-grained spatial relationships (like topology or distance) over long-range, fragmented experiences. To address this, we first propose LASAR, an architecture featuring a dual-memory system designed to maintain both episodic experiences and a semantic cognitive map. We then introduce Spatio-temporal Contextual Representation Learning (ST-CRL), a contrastive objective designed to train this architecture. ST-CRL leverages spatio-temporal cues from cognitive queries generated through annotated spatio-temporal context in simulation to build sample pairs, thereby forming the internal cognitive map from the agent's experiences. Experiments demonstrate that our method achieves 2\%-3.5\% gains in both zero-shot generalization on standard VLN-CE and VSI-Bench benchmarks. We also demonstrate that our proposed cognitive map has high self-consistency.
format Preprint
id arxiv_https___arxiv_org_abs_2605_16899
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle LASAR: Towards Spatio-temporal Reasoning with Latent Cognitive Map
Tang, Jinzhou
Liu, Sidi
Xiu, Waikit
Chen, Weixing
Wang, Keze
Computer Vision and Pattern Recognition
A fundamental challenge in embodied AI is verifying if agents build internal models of spatial structure or merely learn to mimic task-specific expert trajectories. This is critical as foundational approaches rooted in action-centric tasks (e.g., VLN) and reasoning-centric tasks (e.g., EQA) often share a common limitation: they lack a learning signal that forces them to encode fine-grained spatial relationships (like topology or distance) over long-range, fragmented experiences. To address this, we first propose LASAR, an architecture featuring a dual-memory system designed to maintain both episodic experiences and a semantic cognitive map. We then introduce Spatio-temporal Contextual Representation Learning (ST-CRL), a contrastive objective designed to train this architecture. ST-CRL leverages spatio-temporal cues from cognitive queries generated through annotated spatio-temporal context in simulation to build sample pairs, thereby forming the internal cognitive map from the agent's experiences. Experiments demonstrate that our method achieves 2\%-3.5\% gains in both zero-shot generalization on standard VLN-CE and VSI-Bench benchmarks. We also demonstrate that our proposed cognitive map has high self-consistency.
title LASAR: Towards Spatio-temporal Reasoning with Latent Cognitive Map
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2605.16899