Unverified50% confidenceFactExact time
GPTQ was proposed in 2022 and extends the OBS framework from pruning to quantization, using the inverse of the Hessian matrix to optimally adjust unquantized weights to compensate for quantization error, with efficient updates via Cholesky decomposition.
1
Sources
50%
Confidence
Long-term
Relevance
8/7/2026
First Seen
Sources
图解量化：大语言模型压缩核心技术详解
Related Claims
UnverifiedGPTQ（2022）引入逐层Hessian矩阵近似，通过最小化权重量化前后的输出重建误差优化量化参数72% similarUnverifiedThe GPTQ quantization algorithm was proposed by Tim Dettmers et al. in 2023, optimizing quantization error layer by layer through second-order gradient information.70% similarUnverifiedGPTQ通过对每一层权重进行二阶近似优化（基于OBS/OBQ框架）来最小化量化误差，通常需要如C4数据集的128条样本校准，单GPU上仅需数小时完成68% similarUnverifiedModern quantization schemes such as GPTQ, AWQ, and the grouped quantization of the GGUF format can reduce model size by more than 75% while retaining the vast majority of reasoning capability through layer-by-layer calibration and outlier protection.68% similarUnverified现代量化方法如 GPTQ、AWQ 通过逐层校准显著缓解了精度退化问题64% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/705553API
curl https://kongchang.com/api/v1/knowledge/claims/705553MCP
get_claim(id=705553)