# AirLLM: Diffusion Policy-based Adaptive LoRA for Remote Fine-Tuning of LLM over the Air

arXiv: https://arxiv.org/abs/2507.11515 · HTML: https://mlsanity.com/paper/2507.11515/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in Discrete and text diffusion, one of 14 directions in Diffusion models.
- Builds on 25 of its 39 references that are in the map.
- First posted on arXiv 15 Jul 2025.

## Placed in

- Diffusion models / [Discrete and text diffusion](https://mlsanity.com/diffusion/discrete-and-text-diffusion/)

## Builds on (most cited first, 25 of 25)

- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Proximal Policy Optimization Algorithms](https://mlsanity.com/paper/1707.06347/) (2017)
- [LoRA: Low-Rank Adaptation of Large Language Models](https://mlsanity.com/paper/2106.09685/) (2021)
- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [Classifier-Free Diffusion Guidance](https://mlsanity.com/paper/2207.12598/) (2022)
- [OPT: Open Pre-trained Transformer Language Models](https://mlsanity.com/paper/2205.01068/) (2022)
- [Diffusion Policy: Visuomotor Policy Learning via Action Diffusion](https://mlsanity.com/paper/2303.04137/) (2023)
- [A Survey of Large Language Models](https://mlsanity.com/paper/2303.18223/) (2023)
- [Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding](https://mlsanity.com/paper/1510.00149/) (2015)
- [Deep Reinforcement Learning that Matters](https://mlsanity.com/paper/1709.06560/) (2017)
- [Pre-trained Models for Natural Language Processing: A Survey](https://mlsanity.com/paper/2003.08271/) (2020)
- [AdaLoRA: Adaptive Budget Allocation for Parameter-Efficient Fine-Tuning](https://mlsanity.com/paper/2303.10512/) (2023)
- [Diffusion Policy Policy Optimization](https://mlsanity.com/paper/2409.00588/) (2024)
- [Scaling Down to Scale Up: A Guide to Parameter-Efficient Fine-Tuning](https://mlsanity.com/paper/2303.15647/) (2023)
- [Learning Continuous Control Policies by Stochastic Value Gradients](https://mlsanity.com/paper/1510.09142/) (2015)
- [What Matters In On-Policy Reinforcement Learning? A Large-Scale Empirical Study](https://mlsanity.com/paper/2006.05990/) (2020)
- [Wider and Deeper LLM Networks are Fairer LLM Evaluators](https://mlsanity.com/paper/2308.01862/) (2023)
- [Early stopping and non-parametric regression: An optimal data-dependent stopping rule](https://mlsanity.com/paper/1306.3574/) (2013)
- [Latent Space Policies for Hierarchical Reinforcement Learning](https://mlsanity.com/paper/1804.02808/) (2018)
- [PEFT-U: Parameter-Efficient Fine-Tuning for User Personalization](https://mlsanity.com/paper/2407.18078/) (2024)
- [Real-Time Millimeter-Wave MIMO Channel Sounder for Dynamic Directional Measurements](https://mlsanity.com/paper/1807.11921/) (2018)
- [Off-Policy Reinforcement Learning with High Dimensional Reward](https://mlsanity.com/paper/2408.07660/) (2024)
- [Adaptive Sampling and Joint Semantic-Channel Coding under Dynamic Channel Environment](https://mlsanity.com/paper/2502.07236/) (2025)

## Built on by (most cited first, 1 of 1)

- [Empowering Real-World: A Survey on the Technology, Practice, and Evaluation of LLM-driven Industry Agents](https://mlsanity.com/paper/2510.17491/) (2025)

## Abstract (arXiv)

Operating Large Language Models (LLMs) on edge devices is increasingly challenged by limited communication bandwidth and strained computational and memory costs. Thus, cloud-assisted remote fine-tuning becomes indispensable. Nevertheless, existing Low-Rank Adaptation (LoRA) approaches typically employ fixed or heuristic rank configurations, and the subsequent over-the-air transmission of all LoRA parameters could be rather inefficient. To address this limitation, we develop AirLLM, a hierarchical diffusion policy framework for communication-aware LoRA adaptation. Specifically, AirLLM models the rank configuration as a structured action vector that spans all LoRA-inserted projections. To solve the underlying high-dimensional sequential decision-making problem, a Proximal Policy Optimization (PPO) agent generates coarse-grained decisions by jointly observing wireless states and linguistic complexity, which are then refined via Denoising Diffusion Implicit Models (DDIM) to produce high-resolution, task- and channel-adaptive rank vectors. The two modules are optimized alternatively, with the DDIM trained under the Classifier-Free Guidance (CFG) paradigm to maintain alignment with PPO rewards. Experiments under varying signal-to-noise ratios demonstrate that AirLLM consistently enhances fine-tuning performance while significantly reducing transmission costs, highlighting the effectiveness of reinforcement-driven, diffusion-refined rank adaptation for scalable and efficient remote fine-tuning over the air.