Guardado en:
Detalles Bibliográficos
Autores principales: Nie, Lunyiu, Ding, Zhimin, Yu, Kevin, Cheung, Marco, Jermaine, Chris, Chaudhuri, Swarat
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:https://arxiv.org/abs/2504.07247
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866912529289052160
author Nie, Lunyiu
Ding, Zhimin
Yu, Kevin
Cheung, Marco
Jermaine, Chris
Chaudhuri, Swarat
author_facet Nie, Lunyiu
Ding, Zhimin
Yu, Kevin
Cheung, Marco
Jermaine, Chris
Chaudhuri, Swarat
contents The inference-time resource costs of large language and vision models present a growing challenge in production deployments. We propose the use of foundation model programs, i.e., programs that can invoke foundation models with varying resource costs and performance, as an approach to this problem. Specifically, we present a method that translates a task into a program, then learns a policy for resource allocation that, on each input, selects foundation model "backends" for each program module. The policy uses smaller, cheaper backends to handle simpler subtasks, while allowing more complex subtasks to leverage larger, more capable models. We evaluate the method on two new "streaming" visual question-answering tasks in which a system answers a question on a sequence of inputs, receiving ground-truth feedback after each answer. Compared to monolithic multi-modal models, our implementation achieves up to 98% resource savings with minimal accuracy loss, demonstrating its potential for scalable and resource-efficient multi-modal inference.
format Preprint
id arxiv_https___arxiv_org_abs_2504_07247
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Resource-efficient Inference with Foundation Model Programs
Nie, Lunyiu
Ding, Zhimin
Yu, Kevin
Cheung, Marco
Jermaine, Chris
Chaudhuri, Swarat
Machine Learning
The inference-time resource costs of large language and vision models present a growing challenge in production deployments. We propose the use of foundation model programs, i.e., programs that can invoke foundation models with varying resource costs and performance, as an approach to this problem. Specifically, we present a method that translates a task into a program, then learns a policy for resource allocation that, on each input, selects foundation model "backends" for each program module. The policy uses smaller, cheaper backends to handle simpler subtasks, while allowing more complex subtasks to leverage larger, more capable models. We evaluate the method on two new "streaming" visual question-answering tasks in which a system answers a question on a sequence of inputs, receiving ground-truth feedback after each answer. Compared to monolithic multi-modal models, our implementation achieves up to 98% resource savings with minimal accuracy loss, demonstrating its potential for scalable and resource-efficient multi-modal inference.
title Resource-efficient Inference with Foundation Model Programs
topic Machine Learning
url https://arxiv.org/abs/2504.07247