Unverified50% confidenceFactExact time
AWQ由MIT韩松团队于2023年提出，通过保护显著权重，相较朴素Q4量化在相同压缩比下可减少约20%的精度损失
1
Sources
50%
Confidence
Long-term
Relevance
7/23/2026
First Seen
Sources
Related Claims
VerifiedAWQ discovered in 2023 that only about 1% of salient weights have a large impact on accuracy, and reduces quantization loss by protecting this portion of the weights.80% similarUnverifiedAWQ通过分析激活值分布确定每个通道的量化缩放因子，相较朴素INT4量化能显著降低困惑度损失74% similarUnverifiedPost-training quantization solutions such as GPTQ and AWQ can control performance loss to within 1–2% at INT4 precision68% similarVerified4-bit quantization techniques (such as GPTQ, AWQ, and the GGUF format) can reduce model VRAM requirements to 1/4 of the original, but come with a certain degree of accuracy loss68% similarVerifiedAWQ identifies and protects the weights with the greatest influence on activation values, and typically achieves better accuracy than GPTQ at the same compression ratio.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/598685API
curl https://kongchang.com/api/v1/knowledge/claims/598685MCP
get_claim(id=598685)