analiza aiqbench
This research unifies diverse post-training quantization techniques—such as AWQ and GPTQ—under a single mathematical framework centered on gradient-weighted activation sensitivity. By deriving closed-form optimal scaling exponents, the study demonstrates that accounting for token-gradient heterogeneity significantly reduces excess loss in low-bit-width models without requiring retraining. This theoretical consolidation provides a clearer path for optimizing model compression efficiency while maintaining high fidelity.
Activation Sensitivity as a Unifying Principle for Post-Training Quantization
arXiv:2601.11663v2 Announce Type: replace Abstract: Post-training quantization (PTQ) methods for large language models compress pretrained weights to low bit-widths without retraining by optimizing layer-local proxy objectives. Activation-aware scaling methods such as AWQ protect input channels with large activation norms, whereas second-order error-compensation methods such as GPTQ redistribute rounding error using the inverse of the unweighted input covariance matrix. We show that both paradigms -- along with diagonal Fisher quantization, block reconstruction, and Hadamard rotation methods -- are special cases of minimizing a single loss-derived quadratic form governed by gradient-weighted activation sensitivity. From a Gauss-Newton expansion of the task loss at a linear layer $Y = X W^\top$, we prove that under isotropic output curvature the expected excess loss from weight perturbation $\Delta W$ reduces to $\frac{1}{2 d_{\mathrm{out}}}\operatorname{Tr}(\Delta W H_D \Delta W^\top)$, where $H_D = \frac{1}{n} X^\top D X$ weights each calibration token's outer product by its squared output-gradient norm $D_{tt} = \|\nabla_{y_t}\mathcal{L}\|_2^2$. Within this framework, AWQ corresponds to a diagonal approximation of $H_D$ with uniform token weights ($D = I_n$) optimized via channel scaling, while GPTQ corresponds to Cholesky error compensation on $H_D$ under $D = I_n$. We derive closed-form optimal scaling exponents for diagonal sensitivity protection, formulate sensitivity-weighted Cholesky quantization (SAQ-GPTQ) on $H_D$, and prove bounds on token-gradient heterogeneity bias, cross-layer error amplification across residual blocks, and finite-sample calibration concentration. Controlled numerical simulations confirm that under token-gradient heterogeneity, replacing $X^\top X$ with $X^\top D X$ preserves $>99\%$ top-channel saliency recall and reduces INT3/INT4 excess loss by roughly $3\times$ over unweighted GPTQ.
Abstract:Post-training quantization (PTQ) methods for large language models compress pretrained weights to low bit-widths without retraining by optimizing layer-local proxy objectives. Activation-aware scaling methods such as AWQ protect input channels with large activation norms, whereas second-order error-compensation methods such as GPTQ redistribute rounding error using the inverse of the unweighted input covariance matrix. We show that both paradigms -- along with diagonal Fisher quantization, block reconstruction, and Hadamard rotation methods -- are special cases of minimizing a single loss-derived quadratic form governed by gradient-weighted activation sensitivity. From a Gauss-Newton expansion of the task loss at a linear layer $Y = X W^\top$, we prove that under isotropic output curvature the expected excess loss from weight perturbation $\Delta W$ reduces to $\frac{1}{2 d_{\mathrm{out}}}\operatorname{Tr}(\Delta W H_D \Delta W^\top)$, where $H_D = \frac{1}{n} X^\top D X$ weights each calibration token's outer product by its squared output-gradient norm $D_{tt} = \|\nabla_{y_t}\mathcal{L}\|_2^2$. Within this framework, AWQ corresponds to a diagonal approximation of $H_D$ with uniform token weights ($D = I_n$) optimized via channel scaling, while GPTQ corresponds to Cholesky error compensation on $H_D$ under $D = I_n$. We derive closed-form optimal scaling exponents for diagonal sensitivity protection, formulate sensitivity-weighted Cholesky quantization (SAQ-GPTQ) on $H_D$, and prove bounds on token-gradient heterogeneity bias, cross-layer error amplification across residual blocks, and finite-sample calibration concentration. Controlled numerical simulations confirm that under token-gradient heterogeneity, replacing $X^\top X$ with $X^\top D X$ preserves $>99\%$ top-channel saliency recall and reduces INT3/INT4 excess loss by roughly $3\times$ over unweighted GPTQ.
Machine Learning (cs.LG); Artificial Intelligence (cs.AI)
arXiv:2601.11663 [cs.LG]
(or arXiv:2601.11663v2 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2601.11663 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 342 cuvinte
