Revision History
YaRN · 9 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 NTK-by-parts ramp definition and the direction of YaRN's attention temperature (both were reversed), removed the false claims that Llama 3.1 uses YaRN and that its report cites it, corrected misread compute figures (GPU-hours), evaluation datasets and model names in the results, fixed the paper's author affiliations and revision dates, the claim that inference frameworks implement Dynamic YaRN, the ALiBi 'learned bias' description, NTK-aware fine-tuning performance, the LongRoPE extension figures and the LongRoPE2 summary (plus 16 minor fixes)
Version 9 · Nov 4, 2026, 04:33 PM
- Correction: Jeffrey Quesnelle is co-founder and CTO of Nous Research; Dillon Rolnick is CEO as of October 2026
Version 8 · Oct 27, 2026, 10:34 AM
- Improve internal links (2026-10-13)
Version 7 · Oct 13, 2026, 10:21 AM
- Added 6 contextual internal links
Version 6 · Jul 23, 2026, 03:30 PM
- Render formulas with LaTeX math notation
Version 5 · Jul 11, 2026, 06:09 PM