{"ID":6537583,"CreatedAt":"2026-07-14T02:54:43.516908796Z","UpdatedAt":"2026-07-15T03:28:55.185153975Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2607.11359","arxiv_id":"2607.11359","title":"Efficient Tuning Before Low-Bit Post-Training Quantization for Stochastic Gradient Descent-optimized Models","abstract":"Post-training quantization (PTQ) compresses deep neural networks for deployment under limited memory and computational budgets. However, low-bit (i.e., 2-bit or 4-bit) PTQ often suffers from substantial performance degradation. Most existing PTQ methods operate on an unconstrained full-precision (FP) model and primarily address quantization errors through post-hoc reconstruction. We argue that low-bit PTQ accuracy is limited not only by post-quantization error minimization, but also by the quantization-error tolerance of a FP model itself. In this paper, we propose Efficient Tuning Before Quantization (ETBQ), a pre-conditioning tuning stage for Stochastic Gradient Descent (SGD)-optimized models before PTQ. During tuning, the FP model is optimized under perturbations sampled from the error distributions of weight and activation quantization, guiding the model toward a loss-landscape region that is less sensitive to the subsequent PTQ. Unlike QAT, ETBQ does not train a fake-quantized deployment model, which is computationally and memory intensive. Instead, ETBQ outputs a FP model that can be used by any PTQ backend. Experiments on CIFAR-100, Tiny-ImageNet, ImageNet, and Cityscapes provide consistent evidence that ETBQ improves low-bit PTQ across diverse tasks. Under W2A4 settings, e.g., ETBQ improves over naive PTQ by 2.14\\% top-1 accuracy on Tiny-ImageNet and by 5.80\\% mIoU on Cityscapes. Code is available at https://github.com/xpxpxp2001xpxpxp/ETBQ.","short_abstract":"Post-training quantization (PTQ) compresses deep neural networks for deployment under limited memory and computational budgets. However, low-bit (i.e., 2-bit or 4-bit) PTQ often suffers from substantial performance degradation. Most existing PTQ methods operate on an unconstrained full-precision (FP) model and primaril...","url_abs":"https://arxiv.org/abs/2607.11359","url_pdf":"https://arxiv.org/pdf/2607.11359v1","authors":"[\"Peng Xia\",\"Junbiao Pang\",\"Muhammad Ayub Sabir\"]","published":"2026-07-13T10:24:36Z","proceeding":"cs.CV","tasks":"[\"cs.CV\"]","methods":"[]","has_code":false,"code_links":[{"ID":614214,"CreatedAt":"2026-07-14T02:54:43.516908796Z","UpdatedAt":"2026-07-14T02:54:43.516908796Z","DeletedAt":null,"paper_id":6537583,"paper_url":"https://arxiv.org/abs/2607.11359","paper_title":"Efficient Tuning Before Low-Bit Post-Training Quantization for Stochastic Gradient Descent-optimized Models","repo_url":"https://github.com/xpxpxp2001xpxpxp/ETBQ","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}