Revision History
AlpacaEval · 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 null-model cheating result (constant responses, LC win rate 86.5% higher than raw 76.9%, so length control did not help) and its misattributed authors, the false claim that the 805 instructions are private, annotator correlations that measure agreement with human-annotation rankings rather than Chatbot Arena, and MT-Bench's Chatbot Arena correlation (0.94, not lower than both AlpacaEval versions) (plus 13 minor fixes)
Version 9 · Nov 4, 2026, 03:31 PM
- Improve internal links (2026-10-13)
Version 7 · Oct 13, 2026, 10:14 AM
- Added 6 contextual internal links
Version 6 · Jul 23, 2026, 03:27 PM