# Oh! We Freeze: Improving Quantized Knowledge Distillation via Signal Propagation Analysis for Large Language Models

arXiv: https://arxiv.org/abs/2403.18159 · HTML: https://mlsanity.com/paper/2403.18159/

## Key facts

- Cited by 3 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion foundations and sampling, one of 14 directions in Diffusion models.
- Builds on 16 of its 16 references that are in the map.
- First posted on arXiv 26 Mar 2024.

## Placed in

- Diffusion models / [Diffusion foundations and sampling](https://mlsanity.com/diffusion/diffusion-foundations-and-sampling/)
- Knowledge distillation / [Knowledge distillation for LLMs](https://mlsanity.com/distill/knowledge-distillation-for-llms/)

## Builds on (most cited first, 16 of 16)

- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Diffusion Models Beat GANs on Image Synthesis](https://mlsanity.com/paper/2105.05233/) (2021)
- [GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers](https://mlsanity.com/paper/2210.17323/) (2022)
- [Quantization and Training of Neural Networks for Efficient Integer-Arithmetic-Only Inference](https://mlsanity.com/paper/1712.05877/) (2017)
- [Learned Step Size Quantization](https://mlsanity.com/paper/1902.08153/) (2019)
- [Up or Down? Adaptive Rounding for Post-Training Quantization](https://mlsanity.com/paper/2004.10568/) (2020)
- [ZeroQuant: Efficient and Affordable Post-Training Quantization for Large-Scale Transformers](https://mlsanity.com/paper/2206.01861/) (2022)
- [Quantizing deep convolutional networks for efficient inference: A whitepaper](https://mlsanity.com/paper/1806.08342/) (2018)
- [A White Paper on Neural Network Quantization](https://mlsanity.com/paper/2106.08295/) (2021)
- [HAWQ: Hessian AWare Quantization of Neural Networks with Mixed-Precision](https://mlsanity.com/paper/1905.03696/) (2019)
- [Data-Free Quantization Through Weight Equalization and Bias Correction](https://mlsanity.com/paper/1906.04721/) (2019)
- [LLM-QAT: Data-Free Quantization Aware Training for Large Language Models](https://mlsanity.com/paper/2305.17888/) (2023)
- [LSQ+: Improving low-bit quantization through learnable offsets and better initialization](https://mlsanity.com/paper/2004.09576/) (2020)
- [Overcoming Oscillations in Quantization-Aware Training](https://mlsanity.com/paper/2203.11086/) (2022)
- [Neural Network Quantization with AI Model Efficiency Toolkit (AIMET)](https://mlsanity.com/paper/2201.08442/) (2022)

## Built on by (most cited first, 3 of 3)

- [FBI-LLM: Scaling Up Fully Binarized LLMs from Scratch via Autoregressive Distillation](https://mlsanity.com/paper/2407.07093/) (2024)
- [Binary Neural Networks for Large Language Model: A Survey](https://mlsanity.com/paper/2502.19008/) (2025)
- [QMC: Efficient SLM Edge Inference via Outlier-Aware Quantization and Emergent Memories Co-Design](https://mlsanity.com/paper/2601.14549/) (2026)

## Abstract (arXiv)

Large generative models such as large language models (LLMs) and diffusion models have revolutionized the fields of NLP and computer vision respectively. However, their slow inference, high computation and memory requirement makes it challenging to deploy them on edge devices. In this study, we propose a light-weight quantization aware fine tuning technique using knowledge distillation (KD-QAT) to improve the performance of 4-bit weight quantized LLMs using commonly available datasets to realize a popular language use case, on device chat applications. To improve this paradigm of finetuning, as main contributions, we provide insights into stability of KD-QAT by empirically studying the gradient propagation during training to better understand the vulnerabilities of KD-QAT based approaches to low-bit quantization errors. Based on our insights, we propose ov-freeze, a simple technique to stabilize the KD-QAT process. Finally, we experiment with the popular 7B LLaMAv2-Chat model at 4-bit quantization level and demonstrate that ov-freeze results in near floating point precision performance, i.e., less than 0.7% loss of accuracy on Commonsense Reasoning benchmarks.