WinQ: Accelerating Quantization-Aware Training of Language Models Around Saddle Points

• URL: http://arxiv.org/abs/2605.17471v1
• Date: Sun, 17 May 2026 14:20:51 GMT
• Title: WinQ: Accelerating Quantization-Aware Training of Language Models Around Saddle Points
• Abstract summary: Quantization-aware training is widely adopted to quantize language models.<n>The main bottleneck is its slow convergence and early performance plateau.<n>We propose an algorithm called WinQ to accelerate QAT.
• Score: 33.435461430591126
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Quantization-aware training (QAT) is widely adopted to quantize language models by training full-precision weights using gradients from the quantized model. The main bottleneck is its slow convergence and early performance plateau, particularly below 4-bit-widths. While this problem has been observed in prior work, its precise cause remains unclear. In this paper, we analyze the convergence of QAT by estimating the spectrum of the loss-surface Hessians. We find that the weights converge to flat regions around saddle points, where a large fraction of the Hessian eigenvalues are both positive and negative. During training, an increasing fraction of Hessian eigenvalues concentrates around zero, whose magnitude decreases. At lower bit-widths, the magnitude of eigenvalues in the Hessian spectrum is significantly smaller. To mitigate these issues, we propose an algorithm called WinQ to accelerate QAT, which involves: (1) periodically resetting weights to the linear interpolation of full-precision and quantized weights, reducing the distance to the quantization grid and increasing eigenvalue magnitude, and (2) computing gradients of noise-injected weights to regularize the Hessian. Extensive experiments show that WinQ accelerates QAT by up to 4 times across various quantization methods and models. Under the same training cost, WinQ improves state-of-the-art sub-4-bit quantization by up to 8.8%. These results are consistent across 16 settings with different language models, quantization methods, and bit widths.

Related papers

• SINQ: Sinkhorn-Normalized Quantization for Calibration-Free Low-Precision LLM Weights [8.95245917088986]
  Post-training quantization has emerged as the most widely used strategy for deploying large language models at low precision.<n>Current methods show perplexity degradation at bit-widths less than or equal to 4, partly because representing outliers causes precision issues.<n>We introduce SINQ to augment existing post-training quantizers with an additional second-axis scale factor and a fast Sinkhorn-Knopp-style algorithm.
  arXiv  Detail & Related papers  (2025-09-26T21:22:54Z)
• Scaling Law for Quantization-Aware Training [41.782744728992675]
  Quantization-aware training (QAT) reduces model precision while maintaining performance.<n>Existing QAT scaling laws ignore key factors such as the number of training tokens and quantization granularity.<n>This paper proposes a unified scaling law for QAT that models quantization error as a function of model size, training data volume, and quantization group size.
  arXiv  Detail & Related papers  (2025-05-20T12:54:43Z)
• Low-bit Model Quantization for Deep Neural Networks: A Survey [123.89598730307208]
  This article surveys the recent five-year progress towards low-bit quantization on deep neural networks (DNNs)<n>We discuss and compare the state-of-the-art quantization methods and classify them into 8 main categories and 24 sub-categories according to their core techniques.<n>We shed light on the potential research opportunities in the field of model quantization.
  arXiv  Detail & Related papers  (2025-05-08T13:26:19Z)
• PTQ1.61: Push the Real Limit of Extremely Low-Bit Post-Training Quantization Methods for Large Language Models [64.84734437930362]
  Large Language Models (LLMs) suffer severe performance degradation when facing extremely low-bit (sub 2-bit) quantization.<n>We propose an extremely low-bit PTQ method called PTQ1.61, which enables weight quantization to 1.61-bit for the first time.<n>Experiments indicate our PTQ1.61 achieves state-of-the-art performance in extremely low-bit quantization.
  arXiv  Detail & Related papers  (2025-02-18T08:04:58Z)
• ResQ: Mixed-Precision Quantization of Large Language Models with Low-Rank Residuals [10.860081994662645]
  Post-training quantization of large language models (LLMs) holds the promise in reducing the prohibitive computational cost at inference time.<n>We propose ResQ, a PTQ method that pushes further the state-of-the-art.<n>We demonstrate that ResQ outperforms recent uniform and mixed precision PTQ methods on a variety of benchmarks.
  arXiv  Detail & Related papers  (2024-12-18T22:01:55Z)
• GWQ: Gradient-Aware Weight Quantization for Large Language Models [56.22507677736051]
  Large language models (LLMs) show impressive performance in solving complex language tasks.<n> compressing LLMs to low bits can enable to deploy on resource-constrained devices.<n>We propose gradient-aware weight quantization (GWQ), the first quantization approach for low-bit weight quantization.
  arXiv  Detail & Related papers  (2024-10-30T11:16:04Z)
• FlatQuant: Flatness Matters for LLM Quantization [58.28221892035609]
  We propose FlatQuant, a new post-training quantization approach that enhances the flatness of weights and activations.<n>Our approach identifies optimal affine transformations for each linear layer, calibrated in hours via a lightweight objective.<n>It achieves less than 1% accuracy drop for W4A4 quantization on the LLaMA-3-70B model, surpassing SpinQuant by 7.5%.
  arXiv  Detail & Related papers  (2024-10-12T08:10:28Z)
• Mitigating the Impact of Outlier Channels for Language Model Quantization with Activation Regularization [62.15918574997175]
  It is known that language models contain outlier channels whose values on average are orders of magnitude higher than other channels. We propose a strategy which regularizes a layer's inputs via quantization-aware training (QAT) and its outputs via activation kurtosis regularization. We show that regularizing both the inputs and outputs is crucial for preventing a model's "migrating" the difficulty in input quantization to the weights.
  arXiv  Detail & Related papers  (2024-04-04T17:25:30Z)
• Designing strong baselines for ternary neural network quantization through support and mass equalization [7.971065005161565]
  Deep neural networks (DNNs) offer the highest performance in a wide range of applications in computer vision. This computational burden can be dramatically reduced by quantizing floating point values to ternary values. We show experimentally that our approach allows to significantly improve the performance of ternary quantization through a variety of scenarios.
  arXiv  Detail & Related papers  (2023-06-30T07:35:07Z)
• Q-ASR: Integer-only Zero-shot Quantization for Efficient Speech Recognition [65.7040645560855]
  We propose Q-ASR, an integer-only, zero-shot quantization scheme for ASR models. We show negligible WER change as compared to the full-precision baseline models. Q-ASR exhibits a large compression rate of more than 4x with small WER degradation.
  arXiv  Detail & Related papers  (2021-03-31T06:05:40Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.