factual
GPT-4 Elo gain came mostly from post-training
The roughly 100-point Elo score improvement of current GPT-4 over the originally released version came mostly from post-training, driven by multiple stacking axes—data quality, data quantity, more iterations of deploying and collecting new data, and changing annotation types—that together yield a substantial effective compute increase.
factualpending
Evidence Quote
“most of that is post-training. There are a lot of different, separate axes for improvement.”
Source
John Schulman (OpenAI Cofounder) — Reasoning, RLHF, & plan for 2027 AGI— Dwarkesh Patel PodcastCreated: 6/13/2026, 3:34:30 AM
Your notes
Loading notes...