Revision History
Step size · 8 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 direction of Adam's L2 weakness (large, not small, gradients), GPT-3's decay horizon (260B not 300B tokens), Llama 3 8B/70B learning rates and warmup, unsupported Llama 3 weight-decay/clip/beta2 attributions (now cited to GPT-3 and Llama 2), the Smith and Le batch-size/learning-rate noise relationship, and the super-convergence 80k-vs-10k iteration comparison (plus 14 minor fixes)
Version 8 · Nov 4, 2026, 10:46 PM
- Improve internal links (2026-10-13)
Version 7 · Oct 13, 2026, 10:06 AM
- Added 6 contextual internal links
Version 6 · Jul 23, 2026, 03:03 PM
- Render formulas with LaTeX math notation
Version 5 · Jul 11, 2026, 09:28 PM