Saved in:
Bibliographic Details
Main Authors: Bao, Guangsheng, Zhang, Hongbo, Cui, Han, Sun, Ke, Zhao, Yanbin, He, Juncai, Zhang, Yue
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.04651
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866913103450472448
author Bao, Guangsheng
Zhang, Hongbo
Cui, Han
Sun, Ke
Zhao, Yanbin
He, Juncai
Zhang, Yue
author_facet Bao, Guangsheng
Zhang, Hongbo
Cui, Han
Sun, Ke
Zhao, Yanbin
He, Juncai
Zhang, Yue
contents Adapting pretrained models typically involves a trade-off between the high training costs of backpropagation and the heavy inference overhead of memory-based or in-context learning. We propose FAAST, a forward-only associative adaptation method that analytically compiles labeled examples into fast weights in a single pass. By eliminating memory or context dependence, FAAST achieves constant-time inference and decouples task adaptation from pretrained representation. Across image classification and language modeling benchmarks, FAAST matches or exceeds backprop-based adaptation while reducing adaptation time by over 90% and is competitive to memory/context-based adaptation while saving memory usage by up to 95%. These results demonstrate FAAST as a highly efficient, scalable solution for supervised task adaptation, particularly for resource-constrained models. We release the code and models at https://github.com/baoguangsheng/faast.
format Preprint
id arxiv_https___arxiv_org_abs_2605_04651
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle FAAST: Forward-Only Associative Learning via Closed-Form Fast Weights for Test-Time Supervised Adaptation
Bao, Guangsheng
Zhang, Hongbo
Cui, Han
Sun, Ke
Zhao, Yanbin
He, Juncai
Zhang, Yue
Machine Learning
Computation and Language
Adapting pretrained models typically involves a trade-off between the high training costs of backpropagation and the heavy inference overhead of memory-based or in-context learning. We propose FAAST, a forward-only associative adaptation method that analytically compiles labeled examples into fast weights in a single pass. By eliminating memory or context dependence, FAAST achieves constant-time inference and decouples task adaptation from pretrained representation. Across image classification and language modeling benchmarks, FAAST matches or exceeds backprop-based adaptation while reducing adaptation time by over 90% and is competitive to memory/context-based adaptation while saving memory usage by up to 95%. These results demonstrate FAAST as a highly efficient, scalable solution for supervised task adaptation, particularly for resource-constrained models. We release the code and models at https://github.com/baoguangsheng/faast.
title FAAST: Forward-Only Associative Learning via Closed-Form Fast Weights for Test-Time Supervised Adaptation
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2605.04651