Quantization Meets OOD: Generalizable Quantization-aware Training from a Flatness Perspective

• URL: http://arxiv.org/abs/2509.00859v1
• Date: Sun, 31 Aug 2025 14:05:14 GMT
• Title: Quantization Meets OOD: Generalizable Quantization-aware Training from a Flatness Perspective
• Authors: Jiacheng Jiang, Yuan Meng, Chen Tang, Han Yu, Qun Li, Zhi Wang, Wenwu Zhu,
• Abstract summary: We show that quantization-aware training (QAT) methods overlook the potential performance degradation on out-of-distribution (OOD) data.<n>We propose a flatness-oriented QAT method, FQAT, to achieve generalizable QAT.
• Score: 47.71882785472484
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Current quantization-aware training (QAT) methods primarily focus on enhancing the performance of quantized models on in-distribution (I.D) data, while overlooking the potential performance degradation on out-of-distribution (OOD) data. In this paper, we first substantiate this problem through rigorous experiment, showing that QAT can lead to a significant OOD generalization performance degradation. Further, we find the contradiction between the perspective that flatness of loss landscape gives rise to superior OOD generalization and the phenomenon that QAT lead to a sharp loss landscape, can cause the above problem. Therefore, we propose a flatness-oriented QAT method, FQAT, to achieve generalizable QAT. Specifically, i) FQAT introduces a layer-wise freezing mechanism to mitigate the gradient conflict issue between dual optimization objectives (i.e., vanilla QAT and flatness). ii) FQAT proposes an disorder-guided adaptive freezing algorithm to dynamically determines which layers to freeze at each training step, effectively addressing the challenges caused by interference between layers. A gradient disorder metric is designed to help the algorithm identify unstable layers during training. Extensive experiments on influential OOD benchmark demonstrate the superiority of our method over state-of-the-art baselines under both I.D and OOD image classification tasks.

Related papers

• Continual Action Quality Assessment via Adaptive Manifold-Aligned Graph Regularization [53.82400605816587]
  Action Quality Assessment (AQA) quantifies human actions in videos, supporting applications in sports scoring, rehabilitation, and skill evaluation.<n>A major challenge lies in the non-stationary nature of quality distributions in real-world scenarios.<n>We introduce Continual AQA (CAQA), which equips with Continual Learning capabilities to handle evolving distributions.
  arXiv  Detail & Related papers  (2025-10-08T10:09:47Z)
• Sycophancy Mitigation Through Reinforcement Learning with Uncertainty-Aware Adaptive Reasoning Trajectories [58.988535279557546]
  We introduce textbf sycophancy Mitigation through Adaptive Reasoning Trajectories.<n>We show that SMART significantly reduces sycophantic behavior while preserving strong performance on out-of-distribution inputs.
  arXiv  Detail & Related papers  (2025-09-20T17:09:14Z)
• VQAThinker: Exploring Generalizable and Explainable Video Quality Assessment via Reinforcement Learning [50.34205095371895]
  Video quality assessment aims to objectively quantify perceptual quality degradation.<n>Existing VQA models suffer from two critical limitations.<n>We propose textbfVQAThinker, a reasoning-based VQA framework.
  arXiv  Detail & Related papers  (2025-08-08T06:16:23Z)
• Stabilizing Quantization-Aware Training by Implicit-Regularization on Hessian Matrix [0.7261171488281837]
  We find that the sharp landscape of loss, which leads to a dramatic performance drop, is an essential factor that causes instability.<n>We propose Feature-Perturbed Quantization (FPQ) to generalize and employ the feature distillation method to the quantized model.
  arXiv  Detail & Related papers  (2025-03-14T07:56:20Z)
• Directional Gradient Projection for Robust Fine-Tuning of Foundation Models [25.04763038570959]
  Directional Gradient Projection (DiGraP) is a layer-wise trainable method that incorporates directional information from gradients to bridge regularization and multi-objective optimization.<n>We first bridge the uni-modal and multi-modal gap by performing analysis on Image Classification reformulated Visual Question Answering (VQA) benchmarks.<n> Experimental results show that DiGraP consistently outperforms existing baselines across Image Classfication and VQA tasks with discriminative and generative backbones.
  arXiv  Detail & Related papers  (2025-02-21T19:31:55Z)
• DP-IQA: Utilizing Diffusion Prior for Blind Image Quality Assessment in the Wild [73.6767681305851]
  Blind image quality assessment (IQA) in the wild presents significant challenges.<n>Given the difficulty in collecting large-scale training data, leveraging limited data to develop a model with strong generalization remains an open problem.<n>Motivated by the robust image perception capabilities of pre-trained text-to-image (T2I) diffusion models, we propose a novel IQA method, diffusion priors-based IQA.
  arXiv  Detail & Related papers  (2024-05-30T12:32:35Z)
• When No-Reference Image Quality Models Meet MAP Estimation in Diffusion Latents [92.45867913876691]
  No-reference image quality assessment (NR-IQA) models can effectively quantify perceived image quality.<n>We show that NR-IQA models can be plugged into the maximum a posteriori (MAP) estimation framework for image enhancement.
  arXiv  Detail & Related papers  (2024-03-11T03:35:41Z)
• In-Distribution Consistency Regularization Improves the Generalization of Quantization-Aware Training [16.475151881506914]
  We propose Consistency Regularization (CR) to improve the generalization ability of Quantization-Aware Training (QAT)<n>Our approach significantly outperforms current state-of-the-art QAT methods and even the FP counterparts.
  arXiv  Detail & Related papers  (2024-02-21T03:19:48Z)
• You Only Train Once: A Unified Framework for Both Full-Reference and No-Reference Image Quality Assessment [45.62136459502005]
  We propose a network to perform full reference (FR) and no reference (NR) IQA. We first employ an encoder to extract multi-level features from input images. A Hierarchical Attention (HA) module is proposed as a universal adapter for both FR and NR inputs. A Semantic Distortion Aware (SDA) module is proposed to examine feature correlations between shallow and deep layers of the encoder.
  arXiv  Detail & Related papers  (2023-10-14T11:03:04Z)
• Task-Specific Normalization for Continual Learning of Blind Image Quality Models [105.03239956378465]
  We present a simple yet effective continual learning method for blind image quality assessment (BIQA) The key step in our approach is to freeze all convolution filters of a pre-trained deep neural network (DNN) for an explicit promise of stability. We assign each new IQA dataset (i.e., task) a prediction head, and load the corresponding normalization parameters to produce a quality score. The final quality estimate is computed by black a weighted summation of predictions from all heads with a lightweight $K$-means gating mechanism.
  arXiv  Detail & Related papers  (2021-07-28T15:21:01Z)
• Cross Learning in Deep Q-Networks [82.20059754270302]
  We propose a novel cross Q-learning algorithm, aim at alleviating the well-known overestimation problem in value-based reinforcement learning methods. Our algorithm builds on double Q-learning, by maintaining a set of parallel models and estimate the Q-value based on a randomly selected network.
  arXiv  Detail & Related papers  (2020-09-29T04:58:17Z)
• Norm-in-Norm Loss with Faster Convergence and Better Performance for Image Quality Assessment [20.288424566444224]
  We explore normalization in the design of loss functions for image quality assessment (IQA) models. The resulting "Norm-in-Norm'' loss encourages the IQA model to make linear predictions with respect to subjective quality scores. Experiments on two relevant datasets show that, compared to MAE or MSE loss, the new loss enables the IQA model to converge about 10 times faster.
  arXiv  Detail & Related papers  (2020-08-10T04:01:21Z)

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.