Dwarkesh Pretraining Replication Finds Data Gains Deliver 3.24x More Compute Multipliers Than Model Gains
josh_wills · x · 2026-09-09
Dwarkesh Patel and collaborator @whoisjerbear pretrained combinations of year-representative open model recipes and data corpora from 2019–2025 at various small scales. Key findings:
- Data improvements contributed 3.24x as many compute multipliers as model improvements (12.0x vs 3.7x), suggesting pretraining progress now comes mostly from data.
- The gains stack independently: a better dataset helps every architecture about equally, and vice versa.
Full results and implications for future AI progress are published in the linked writeup.
Related event: Dwarkesh Experiments: Data Drives Most Pretraining Progress(3 posts)→
More from Research
- DeepMind paper: LLMs can't sample randomly on their own — pass random numbers from code instead — rohanpaul_ai · 2026-10-27
- World Model Assigns Weight and Friction to Objects, Then Simulates with a Physics Engine — jonstephens85 · 2026-10-27
- Token Collapse: Study Finds Graph Soft Prompting Tokens in LLMs Fail to Encode Structure — Sapienza · 2026-10-27
- Mathathon 2.0 launches with Terence Tao's SAIR, $1.5M in compute for teams — _sathvikr · 2026-10-27
- The 'sweet lesson' is real: two kinds of ML problems need opposite strategies — mike64_t · 2026-10-27
- PixGR: Pixel-space generative decoding for fast feed-forward 3D reconstruction — kwangmoo_yi · 2026-10-27