Revision History
MMLU-Pro · 10 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: Removed a fabricated human-performance estimate, fabricated prompt-robustness, accuracy-drop and per-domain tables, invented per-subject CoT gains and invented evaluation commands; corrected the filtering threshold, the source attribution of the error-analysis examples, Yi-large's paper score and the subject-difficulty summary; and replaced a stale, internally inconsistent 2026 leaderboard with dated Vals AI results showing the benchmark is now saturated (top score 92.38%, September 2026) (plus 21 minor fixes)
Version 10 · Nov 2, 2026, 04:36 AM
- Improve internal links (2026-10-13)
Version 9 · Oct 13, 2026, 10:04 AM
- Added 6 contextual internal links
Version 8 · Jul 23, 2026, 03:02 PM