Nonlinear Bipolar Compensation: Handling Outliers in Post-Training Quantization

• URL: http://arxiv.org/abs/2605.16423v1
• Date: Thu, 14 May 2026 14:55:46 GMT
• Title: Nonlinear Bipolar Compensation: Handling Outliers in Post-Training Quantization
• Abstract summary: Network quantization significantly reduces a model's memory and compute consumption by mapping floating-point numbers to low-bit representations.<n>Existing quantization methods typically suffer from the speed-accuracy tradeoff and limited generalization.<n>We propose Bipolar Compensation (NBC), a post-training quantization approach that introduces nonlinear compensation to reduce the effect of outliers.
• Score: 11.76317251442695
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Network quantization has emerged as one of the most practical model compression techniques, which significantly reduces a model's memory and compute consumption by mapping floating-point numbers to low-bit representations. However, existing quantization methods typically suffer from the speed-accuracy tradeoff and limited generalization. To address these issues, recent compensation-based methods offer an efficient yet general solution by introducing additional lightweight linear layers into the quantized network. However, the accuracy of these methods suffers from their limited compensation capability and high sensitivity to outliers. In this paper, we propose Nonlinear Bipolar Compensation (NBC), a post-training quantization approach that introduces nonlinear compensation to reduce the effect of outliers. We further design Bipolar Logarithmic Transformation (BLT), which compresses outliers by mapping both the quantized input and the quantization error into a transformed space. A simple linear layer is then applied for compensation in the transformed space, preserving the efficiency of our method. Extensive experiments across various tasks, models, and quantization methods confirm the effectiveness, efficiency, robustness, and generality of our NBC approach.

Related papers

• WUSH: Near-Optimal Adaptive Transforms for LLM Quantization [52.77441224845925]
  Quantization to low bitwidth is a standard approach for deploying large language models.<n>A few extreme weights and activations stretch the dynamic range and reduce the effective resolution of the quantizer.<n>We derive, for the first time, closed-form optimal linear blockwise transforms for joint weight-activation quantization.
  arXiv  Detail & Related papers  (2025-11-30T16:17:34Z)
• Pushing the Limits of Large Language Model Quantization via the Linearity Theorem [71.3332971315821]
  We present a "line theoremarity" establishing a direct relationship between the layer-wise $ell$ reconstruction error and the model perplexity increase due to quantization. This insight enables two novel applications: (1) a simple data-free LLM quantization method using Hadamard rotations and MSE-optimal grids, dubbed HIGGS, and (2) an optimal solution to the problem of finding non-uniform per-layer quantization levels.
  arXiv  Detail & Related papers  (2024-11-26T15:35:44Z)
• Robust Training of Neural Networks at Arbitrary Precision and Sparsity [15.121043556313689]
  We introduce a denoising dequantization transform derived from a principled ridge regression objective.<n>We extend this principle to sparsification by viewing it as a special form of quantization that maps insignificant values to zero.<n>This approach yields state-of-the-art results and provides a theoretically-grounded path to hyper-efficient neural networks.
  arXiv  Detail & Related papers  (2024-09-14T00:57:32Z)
• QuEST: Low-bit Diffusion Model Quantization via Efficient Selective Finetuning [52.157939524815866]
  In this paper, we identify imbalanced activation distributions as a primary source of quantization difficulty.<n>We propose to adjust these distributions through weight finetuning to be more quantization-friendly.<n>Our method demonstrates its efficacy across three high-resolution image generation tasks.
  arXiv  Detail & Related papers  (2024-02-06T03:39:44Z)
• An Accelerated Doubly Stochastic Gradient Method with Faster Explicit Model Identification [97.28167655721766]
  We propose a novel doubly accelerated gradient descent (ADSGD) method for sparsity regularized loss minimization problems. We first prove that ADSGD can achieve a linear convergence rate and lower overall computational complexity.
  arXiv  Detail & Related papers  (2022-08-11T22:27:22Z)
• BiTAT: Neural Network Binarization with Task-dependent Aggregated Transformation [116.26521375592759]
  Quantization aims to transform high-precision weights and activations of a given neural network into low-precision weights/activations for reduced memory usage and computation. Extreme quantization (1-bit weight/1-bit activations) of compactly-designed backbone architectures results in severe performance degeneration. This paper proposes a novel Quantization-Aware Training (QAT) method that can effectively alleviate performance degeneration.
  arXiv  Detail & Related papers  (2022-07-04T13:25:49Z)
• Towards Mixed-Precision Quantization of Neural Networks via Constrained Optimization [28.76708310896311]
  We present a principled framework to solve the mixed-precision quantization problem. We show that our method is derived in a principled way and much more computationally efficient.
  arXiv  Detail & Related papers  (2021-10-13T08:09:26Z)
• Quantized Adam with Error Feedback [11.91306069500983]
  We present a distributed variant of adaptive gradient method for training deep neural networks. We incorporate two types of quantization schemes to reduce the communication cost among the workers.
  arXiv  Detail & Related papers  (2020-04-29T13:21:54Z)
• Gradient $\ell_1$ Regularization for Quantization Robustness [70.39776106458858]
  We derive a simple regularization scheme that improves robustness against post-training quantization. By training quantization-ready networks, our approach enables storing a single set of weights that can be quantized on-demand to different bit-widths.
  arXiv  Detail & Related papers  (2020-02-18T12:31:34Z)

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.