Gradient Transformer: Learning to Generate Updates for LLMs

Binh-Nguyen Nguyen, Khang Tran, NhatHai Phan, Issa Khalil · 2026

Read Gradient Transformer: Learning to Generate Updates for LLMs by Binh-Nguyen Nguyen, Khang Tran, NhatHai Phan, Issa Khalil (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-27591.txt.

Description

Binh-Nguyen Nguyen, Khang Tran, NhatHai Phan, and Issa Khalil generate large-model update vectors from tiny models tuned on private data. Their Gradient Transformer learns the correlation between tiny-model and large-model update vectors from shadow datasets, letting providers update models without touching private data. The approach supports multi-organization collaboration and beats leading distillation baselines even under strict differential privacy.

Catalog record

CREATORBinh-Nguyen Nguyen, Khang Tran, NhatHai Phan, Issa Khalil
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE75998
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.27591

Browse the full library · About the Tower of Babel