Revision History
AWQ (Activation-aware Weight Quantization) · 6 revisions
Sizes are character counts of the article source. The signed number is the change from the previous revision.
Recent edit summaries
Detailed summaries recorded by editors. Generic maintenance summaries are omitted here; every recorded revision remains below. Dates describe the edit, not necessarily the event it covers.
- Correction: Corrected the salient-channel ablation and 3-bit results, which were attributed to LLaMA-7B but come from OPT models (actual LLaMA-7B INT3: AWQ 6.35 vs RTN 7.01), fixed the LLaMA-13B AWQ perplexity (5.19), and corrected GPTQ's calibration set size (128 segments of 2,048 tokens, not 2,048 samples) (plus 19 minor fixes)
Version 7 · Nov 4, 2026, 11:37 PM
- Improve internal links (2026-10-13)
Version 6 · Oct 13, 2026, 10:20 AM
- Added 6 contextual internal links
Version 5 · Jul 23, 2026, 03:29 PM