Revision History
Direct Preference Optimization (DPO) · 11 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 a fabricated author list for the SPPO reference and a fabricated title for the Rashidinejad and Tian reward-hacking reference, and re-sourced the DPO overoptimization finding to the paper that reports it (removing an unsupported claim that the implicit reward rises monotonically) (plus 19 minor fixes)
Version 11 · Nov 2, 2026, 05:38 AM
- Update: added a video.
Version 10 · Nov 1, 2026, 08:25 AM
- Improve internal links (2026-10-13)
Version 9 · Oct 13, 2026, 10:15 AM
- Added 6 contextual internal links
Version 8 · Jul 23, 2026, 03:27 PM