# Gradient Transformer: Learning to Generate Updates for LLMs

arXiv: https://arxiv.org/abs/2605.27591 · HTML: https://mlsanity.com/paper/2605.27591/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Placed in Knowledge distillation for LLMs, one of 12 directions in Knowledge distillation.
- Builds on 32 of its 54 references that are in the map.
- First posted on arXiv 26 May 2026.

## Placed in

- Knowledge distillation / [Knowledge distillation for LLMs](https://mlsanity.com/distill/knowledge-distillation-for-llms/)

## Builds on (most cited first, 30 of 32)

- [Distilling the Knowledge in a Neural Network](https://mlsanity.com/paper/1503.02531/) (2015)
- [Training Verifiers to Solve Math Word Problems](https://mlsanity.com/paper/2110.14168/) (2021)
- [Scaling Laws for Neural Language Models](https://mlsanity.com/paper/2001.08361/) (2020)
- [Scaling Instruction-Finetuned Language Models](https://mlsanity.com/paper/2210.11416/) (2022)
- [Deep Learning with Differential Privacy](https://mlsanity.com/paper/1607.00133/) (2016)
- [CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledge](https://mlsanity.com/paper/1811.00937/) (2018)
- [DROP: A Reading Comprehension Benchmark Requiring Discrete Reasoning Over Paragraphs](https://mlsanity.com/paper/1903.00161/) (2019)
- [Program Induction by Rationale Generation : Learning to Solve and Explain Algebraic Word Problems](https://mlsanity.com/paper/1705.04146/) (2017)
- [Mixture-of-Experts with Expert Choice Routing](https://mlsanity.com/paper/2202.09368/) (2022)
- [Generating Sequences With Recurrent Neural Networks](https://mlsanity.com/paper/1308.0850/) (2013)
- [SAMSum Corpus: A Human-annotated Dialogue Dataset for Abstractive Summarization](https://mlsanity.com/paper/1911.12237/) (2019)
- [Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision](https://mlsanity.com/paper/2312.09390/) (2023)
- [DialogSum: A Real-Life Scenario Dialogue Summarization Dataset](https://mlsanity.com/paper/2105.06762/) (2021)
- [Data-Free Knowledge Distillation for Deep Neural Networks](https://mlsanity.com/paper/1710.07535/) (2017)
- [Asymmetry in Low-Rank Adapters of Foundation Models](https://mlsanity.com/paper/2402.16842/) (2024)
- [Theoretical Analysis of Weak-to-Strong Generalization](https://mlsanity.com/paper/2405.16043/) (2024)
- [A Unified Framework for Quantifying Privacy Risk in Synthetic Data](https://mlsanity.com/paper/2211.10459/) (2022)
- [Generalization Bounds For Meta-Learning: An Information-Theoretic Analysis](https://mlsanity.com/paper/2109.14595/) (2021)
- [Vision Superalignment: Weak-to-Strong Generalization for Vision Foundation Models](https://mlsanity.com/paper/2402.03749/) (2024)
- [Fine Tuning LLM for Enterprise: Practical Guidelines and Recommendations](https://mlsanity.com/paper/2404.10779/) (2024)
- [A Linear Reconstruction Approach for Attribute Inference Attacks against Synthetic Data](https://mlsanity.com/paper/2301.10053/) (2023)
- [A Survey: Collaborative Hardware and Software Design in the Era of Large Language Models](https://mlsanity.com/paper/2410.07265/) (2024)
- [NAYER: Noisy Layer Data Generation for Efficient and Effective Data-free Knowledge Distillation](https://mlsanity.com/paper/2310.00258/) (2023)
- [Weak-to-Strong Generalization Through the Data-Centric Lens](https://mlsanity.com/paper/2412.03881/) (2024)
- [Small Scale Data-Free Knowledge Distillation](https://mlsanity.com/paper/2406.07876/) (2024)
- [Distribution-Dependent Analysis of Gibbs-ERM Principle](https://mlsanity.com/paper/1902.01846/) (2019)
- [Data-free Knowledge Distillation with Diffusion Models](https://mlsanity.com/paper/2504.00870/) (2025)
- [Characterizing and Understanding the Generalization Error of Transfer Learning with Gibbs Algorithm](https://mlsanity.com/paper/2111.01635/) (2021)
- [Discrepancies are Virtue: Weak-to-Strong Generalization through Lens of Intrinsic Dimension](https://mlsanity.com/paper/2502.05075/) (2025)
- [On the Generalization for Transfer Learning: An Information-Theoretic Analysis](https://mlsanity.com/paper/2207.05377/) (2022)

## Abstract (arXiv)

Many organizations lack computational resources to fine-tune large language models (LLMs) on private (unshareable) data for better utility, while fine-tuning tiny language models (TinyLMs) alone performs poorly. To address this bottleneck, we propose a data-free knowledge distillation framework that generates LLM update vectors based on TinyLMs fine-tuned on private data. An update vector is a vector of parameter changes from an initial model to its fine-tuned version on a dataset, capturing the effect of cumulative gradient steps during fine-tuning. The key idea of our framework is a novel Gradient Transformer that transforms TinyLM's update vectors into LLM's update vectors. As derived from shadow datasets, Grad-Transformer captures the correlation between TinyLM and LLM update vectors, enabling third-party providers to generate LLM update vectors given the organization's TinyLM update vectors without accessing the organization's private data. The framework supports multi-organization collaboration to jointly update LLMs, improving performance and cost-efficiency. Extensive experiments across language modeling and reasoning tasks show that Grad-Transformer remarkably outperforms state-of-the-art knowledge distillation baselines, even under strict differential privacy protection.