Combines scaling laws with firm-level profit maximization to derive optimal model size and training expenditure in compute-bound and data-bound regimes.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
A Theory of Training Profit-Optimal LLMs
Combines scaling laws with firm-level profit maximization to derive optimal model size and training expenditure in compute-bound and data-bound regimes.