_version_ 1866910282275618816
author Lab, Mind
:
Cao, Song
Cao, Vic
Chen, Kaijie
Fan, Bunny
Feng, Hera
Feng, Huan
Fu, Arthur
Gao, Jun
Gu, Hongquan
Guan, Aaron
Hong, Mutian
Hou, Hailee
Hua, Peixuan
Huang, Charles
Jiang, Miles
Jiang, Nora
Jiang, Yuyi
Jin, Autumn
Kong, Fancy
Lei, Kyrie
Li, Alexy
Li, Dawn
Li, Ray
Li, Theo
Li, Wenhao
Lin, Jiayi
Liu, Domini
Liu, Heshan
Liu, Kairus
Liu, Logan
Luo, Maeve
Lv, Runism
Ma, Pony
Niu, Verity
Qiu, Anson
Wang, Vincent
Yao, Maxwell
Ye, Regis
Ye, Wenlin
Ye, Yanying
Ying, Josh
Zeng, Danney
Zhan, Salmon
Zhang, Anya
Zhang, Ruijia
Zhang, Shiyang
Zhang, Sueky
Zhang, Ya
Zhao, Wei
Zhou, Ada
Zhou, Sizer
Zhu, Xinyue
Zhuang, Murphy
author_facet Lab, Mind
:
Cao, Song
Cao, Vic
Chen, Kaijie
Fan, Bunny
Feng, Hera
Feng, Huan
Fu, Arthur
Gao, Jun
Gu, Hongquan
Guan, Aaron
Hong, Mutian
Hou, Hailee
Hua, Peixuan
Huang, Charles
Jiang, Miles
Jiang, Nora
Jiang, Yuyi
Jin, Autumn
Kong, Fancy
Lei, Kyrie
Li, Alexy
Li, Dawn
Li, Ray
Li, Theo
Li, Wenhao
Lin, Jiayi
Liu, Domini
Liu, Heshan
Liu, Kairus
Liu, Logan
Luo, Maeve
Lv, Runism
Ma, Pony
Niu, Verity
Qiu, Anson
Wang, Vincent
Yao, Maxwell
Ye, Regis
Ye, Wenlin
Ye, Yanying
Ying, Josh
Zeng, Danney
Zhan, Salmon
Zhang, Anya
Zhang, Ruijia
Zhang, Shiyang
Zhang, Sueky
Zhang, Ya
Zhao, Wei
Zhou, Ada
Zhou, Sizer
Zhu, Xinyue
Zhuang, Murphy
contents Parameter-efficient fine-tuning (PEFT) is usually treated as a cheaper alternative to full fine-tuning. We study a broader role: small trainable adapters as persistent local state on top of strong shared foundation models. In this framing, the base model provides shared competence while adapters carry instance-specific behavior such as preferences, skills, tool habits, and memory-like updates. We organize the problem around three scaling axes: Scale Up, where stronger shared priors make small local updates more useful; Scale Down, where we study how small adapters can be while remaining reliable; and Scale Out, where many persistent adapted instances coexist. MinT provides one infrastructure example for managing adapter identity, revision, provenance, evaluation, and serving residency. Together, the results suggest that PEFT can be a compact substrate for persistent personal models rather than only a budget substitute for full fine-tuning.
format Preprint
id arxiv_https___arxiv_org_abs_2606_02437
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle On the Scaling of PEFT: Towards Million Personal Models of Trillion Parameters
Lab, Mind
:
Cao, Song
Cao, Vic
Chen, Kaijie
Fan, Bunny
Feng, Hera
Feng, Huan
Fu, Arthur
Gao, Jun
Gu, Hongquan
Guan, Aaron
Hong, Mutian
Hou, Hailee
Hua, Peixuan
Huang, Charles
Jiang, Miles
Jiang, Nora
Jiang, Yuyi
Jin, Autumn
Kong, Fancy
Lei, Kyrie
Li, Alexy
Li, Dawn
Li, Ray
Li, Theo
Li, Wenhao
Lin, Jiayi
Liu, Domini
Liu, Heshan
Liu, Kairus
Liu, Logan
Luo, Maeve
Lv, Runism
Ma, Pony
Niu, Verity
Qiu, Anson
Wang, Vincent
Yao, Maxwell
Ye, Regis
Ye, Wenlin
Ye, Yanying
Ying, Josh
Zeng, Danney
Zhan, Salmon
Zhang, Anya
Zhang, Ruijia
Zhang, Shiyang
Zhang, Sueky
Zhang, Ya
Zhao, Wei
Zhou, Ada
Zhou, Sizer
Zhu, Xinyue
Zhuang, Murphy
Machine Learning
Computation and Language
Parameter-efficient fine-tuning (PEFT) is usually treated as a cheaper alternative to full fine-tuning. We study a broader role: small trainable adapters as persistent local state on top of strong shared foundation models. In this framing, the base model provides shared competence while adapters carry instance-specific behavior such as preferences, skills, tool habits, and memory-like updates. We organize the problem around three scaling axes: Scale Up, where stronger shared priors make small local updates more useful; Scale Down, where we study how small adapters can be while remaining reliable; and Scale Out, where many persistent adapted instances coexist. MinT provides one infrastructure example for managing adapter identity, revision, provenance, evaluation, and serving residency. Together, the results suggest that PEFT can be a compact substrate for persistent personal models rather than only a budget substitute for full fine-tuning.
title On the Scaling of PEFT: Towards Million Personal Models of Trillion Parameters
topic Machine Learning
Computation and Language
url https://arxiv.org/abs/2606.02437