Spend Less, Fit Better: Budget-Efficient Scaling Law Fitting via Active Experiment Selection
Sijie Li, Shanda Li, Haowei Lin, Weiwei Sun, Ameet Talwalkar, Yiming Yang, [0.5em] Peking University · 2026
Read Spend Less, Fit Better: Budget-Efficient Scaling Law Fitting via Active Experiment Selection by Sijie Li, Shanda Li, Haowei Lin, Weiwei Sun, Ameet Talwalkar, Yiming Yang, [0.5em] Peking University (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-22753.txt.
Description
Li and five colleagues treat scaling-law fitting as budget-aware experimental design, choosing which pilot runs to execute so extrapolation into a costly target region stays accurate. Their SL-squared method allocates budget sequentially toward runs most useful for that region, guided by uncertainty. Across diverse scaling-law tasks it often nears full-dataset fitting quality while spending only about a tenth of the total training budget.
Catalog record
| CREATOR | Sijie Li, Shanda Li, Haowei Lin, Weiwei Sun, Ameet Talwalkar, Yiming Yang, [0.5em] Peking University |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 72633 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2604.22753