LCLA: Language-Conditioned Latent Alignment for Vision-Language Navigation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Subedi, Nitesh, Haroon, Adam, Tetteh, Samuel, Koirala, Prajwal, Fleming, Cody, Sarkar, Soumik
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917262562164736
author Subedi, Nitesh
Haroon, Adam
Tetteh, Samuel
Koirala, Prajwal
Fleming, Cody
Sarkar, Soumik
author_facet Subedi, Nitesh
Haroon, Adam
Tetteh, Samuel
Koirala, Prajwal
Fleming, Cody
Sarkar, Soumik
contents We propose LCLA (Language-Conditioned Latent Alignment), a framework for vision-language navigation that learns modular perception-action interfaces by aligning sensory observations to a latent representation of an expert policy. The expert is first trained with privileged state information, inducing a latent space sufficient for control, after which its latent interface and action head are frozen. A lightweight adapter is then trained to map raw visual-language observations, via a frozen vision-language model, into the expert's latent space, reducing the problem of visuomotor learning to supervised latent alignment rather than end-to-end policy optimization. This decoupling enforces a stable contract between perception and control, enabling expert behavior to be reused across sensing modalities and environmental variations. We instantiate LCLA and evaluate it on a vision-language indoor navigation task, where aligned latent spaces yield strong in-distribution performance and robust zero-shot generalization to unseen environments, lighting conditions, and viewpoints while remaining lightweight at inference time.
format Preprint
id arxiv_https___arxiv_org_abs_2602_07629
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle LCLA: Language-Conditioned Latent Alignment for Vision-Language Navigation
Subedi, Nitesh
Haroon, Adam
Tetteh, Samuel
Koirala, Prajwal
Fleming, Cody
Sarkar, Soumik
Robotics
We propose LCLA (Language-Conditioned Latent Alignment), a framework for vision-language navigation that learns modular perception-action interfaces by aligning sensory observations to a latent representation of an expert policy. The expert is first trained with privileged state information, inducing a latent space sufficient for control, after which its latent interface and action head are frozen. A lightweight adapter is then trained to map raw visual-language observations, via a frozen vision-language model, into the expert's latent space, reducing the problem of visuomotor learning to supervised latent alignment rather than end-to-end policy optimization. This decoupling enforces a stable contract between perception and control, enabling expert behavior to be reused across sensing modalities and environmental variations. We instantiate LCLA and evaluate it on a vision-language indoor navigation task, where aligned latent spaces yield strong in-distribution performance and robust zero-shot generalization to unseen environments, lighting conditions, and viewpoints while remaining lightweight at inference time.
title LCLA: Language-Conditioned Latent Alignment for Vision-Language Navigation
topic Robotics
url https://arxiv.org/abs/2602.07629