QUASAR: Lowering the Loss Floor of Quantization-Aware Training with Loss-Aware Reconstruction

📄 arXiv: 2608.13966v1 📥 PDF

作者: Vincent Counathe, Ben Athiwaratkun, Christopher De Sa, Tianyi Zhang

分类: cs.LG, cs.CL, stat.ML

发布日期: 2026-08-14

备注: 39 pages

💡 一句话要点

提出QUASAR以降低量化感知训练的损失底线

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 量化感知训练 损失重构 深度学习 模型压缩 低比特推理 大语言模型 优化算法

📋 核心要点

  1. 现有的量化感知训练方法在计算损失和梯度时使用的重构方式存在不匹配，导致训练效果不佳。
  2. QUASAR通过在训练过程中持续进行轻量级的损失感知重构，优化了量化模型的训练过程，降低了损失底线。
  3. 在Qwen3和Llama-3.1模型上，QUASAR在2、3和4比特的量化中实现了最低的KL散度，且在多个任务上提高了准确率。

📝 摘要（中文）

随着大型语言模型推理向低精度转变，后训练量化（PTQ）变得越来越脆弱，因此量化感知训练（QAT）对于保持模型质量至关重要。然而，QAT在使用损失重构的情况下计算损失和替代梯度，这种不匹配可能导致次优的训练轨迹和更高的损失底线。QUASAR是一种新的QAT方法，通过在训练循环中持续执行轻量级的损失感知重构，降低损失底线并改善低比特模型的性能。QUASAR在每个训练步骤中使用平方梯度的指数移动平均作为在线显著性估计，并通过显著性加权最小二乘法拟合仿射去量化器。实验表明，QUASAR在多个基准测试中显著提升了模型性能。

🔬 方法详解

问题定义：论文要解决的问题是现有量化感知训练（QAT）方法在使用损失重构计算梯度时存在的不匹配，导致训练效果不理想和损失底线过高。

核心思路：QUASAR的核心思路是通过在训练循环中持续进行轻量级的损失感知重构，实时调整模型参数，从而降低损失底线并提高量化模型的性能。

技术框架：QUASAR的整体架构包括三个主要模块：损失感知重构模块、显著性估计模块和仿射去量化器拟合模块。每个训练步骤中，QUASAR首先计算梯度的指数移动平均，然后在小范围内搜索剪切范围，最后通过显著性加权最小二乘法拟合去量化器。

关键创新：QUASAR的最重要技术创新在于其持续的损失感知重构过程，这与传统方法在训练过程中只进行一次重构的方式有本质区别，从而显著提高了模型的训练效率和最终性能。

关键设计：QUASAR在损失函数设计上采用了损失感知重构误差作为优化目标，并通过显著性加权的方式来调整去量化器的参数设置，确保了模型在量化后的性能提升。具体参数设置和网络结构细节在实验部分进行了详细描述。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

QUASAR在Qwen3和Llama-3.1模型上表现出色，在2、3和4比特的量化中，KL散度分别降低了至少10%（3和4比特）和29%（2比特）。此外，在2比特的情况下，QUASAR在八个任务上的平均准确率提高了3.5-4.3个百分点，显著优于强基线的QAT和PTQ方法。

🎯 应用场景

QUASAR的研究成果在低精度推理的场景中具有广泛的应用潜力，尤其是在大型语言模型和深度学习模型的部署中。通过降低量化带来的性能损失，QUASAR能够帮助开发者在资源受限的环境中实现高效的模型推理，推动智能应用的普及和发展。

📄 摘要（原文）

As large language model inference shifts toward lower precision, post-training quantization (PTQ) becomes increasingly brittle, making quantization-aware training (QAT) essential for preserving model quality. However, QAT computes the loss and surrogate gradients using a lossy reconstruction of latent full-precision weights, while applying updates to the latent weights themselves. This mismatch can lead to suboptimal training trajectories and a higher loss floor. Second-order PTQ methods mitigate a similar gap by minimizing loss-aware reconstruction error, but doing it once for a frozen model can take hours; repeating this process throughout QAT as the weights evolve is impractical. We introduce QUASAR, a QAT method that continuously performs lightweight, loss-aware reconstruction in the training loop to lower the loss floor and improve the resulting low-bit model. At each training step, QUASAR uses the exponential moving average of squared gradients as online saliency estimates, searches over a small set of clipping ranges, and fits affine dequantizers via saliency-weighted least squares. Our analysis shows that the loss-aware reconstruction error is the only reconstruction-dependent term in the QAT convergence bound and controls the loss of the final quantized model, establishing QUASAR's objective as a principled optimization target. QUASAR modifies only the training procedure and supports standard deployment formats, including integer quantization and NVFP4, with no inference-time changes or overhead. Across Qwen3 and Llama-3.1, QUASAR achieves the lowest held-out KL divergence among competitive QAT methods at 2, 3, and 4 bits, reducing KL by at least 10% at 3 and 4 bits and by 29% at 2 bits. At 2 bits, it improves average accuracy across eight tasks by 3.5-4.3 percentage points over strong QAT and PTQ baselines.