inference/research
verifiedhigh · 0.90updated 2026-10-03

AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration

paper

Fields

id
sources/quant-awq-paper
Publisher
arXiv
Kind
paper
Published
2023-06-01
Accessed
2026-10-03
notes
v6 revised 2026-04-25. MLsys 2024. INT3-g128 / INT4-g128 WikiText perplexity tables. | SOURCE AUDIT 2026-10-03: CONSOLIDATED: A second source record (qlab-awq-paper) pointed at this same arXiv 2306.00978 URL and was cited by nothing; its notes read: "Read to pin down AWQ's activation handling; the record itself already exists in the repo and was not duplicated." That was correct - AWQ is a weight-only scheme and no record claims activation quantization for it - so the redundant record was deleted and this line is its surviving trace.

Notes

v6 revised 2026-04-25. MLsys 2024. INT3-g128 / INT4-g128 WikiText perplexity tables. | SOURCE AUDIT 2026-10-03: CONSOLIDATED: A second source record (qlab-awq-paper) pointed at this same arXiv 2306.00978 URL and was cited by nothing; its notes read: "Read to pin down AWQ's activation handling; the record itself already exists in the repo and was not duplicated." That was correct - AWQ is a weight-only scheme and no record claims activation quantization for it - so the redundant record was deleted and this line is its surviving trace.

Cross-references

This record references no other records.