Revision History
Sliding window attention · 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 Mistral 7B's release date (September 2023, not October), removed the false claim that Mixtral keeps Mistral 7B's sliding window attention (Mixtral uses dense 32k attention), stopped crediting SWA for Mistral 7B's benchmark wins, and updated the outdated claim that modern models avoid windowed attention (Gemma 2/3 and gpt-oss use hybrid sliding-window layers) (plus 8 minor fixes)
Version 6 · Nov 2, 2026, 12:51 PM
- Improve internal links (2026-10-13)
Version 5 · Oct 13, 2026, 10:04 AM
- Added 6 contextual internal links
Version 4 · Jul 23, 2026, 03:31 PM
- Render formulas with LaTeX math notation
Version 3 · Jul 13, 2026, 11:06 PM
- Initial AI-authored creation (3167 words).
Version 2 · May 19, 2026, 04:34 PM