A Scaling Law for Token Efficiency in LLM Fine-Tuning Under Fixed Compute Budgets
This work addresses practical LLM fine-tuning in resource-constrained settings, though it appears incremental as it refines existing scaling law approaches.
The paper tackles the problem of fine-tuning large language models under fixed compute budgets by showing that data composition (number of examples and average token length) significantly affects token efficiency, with experiments on BRICC and MMLU datasets revealing these effects.
We introduce a scaling law for fine-tuning large language models (LLMs) under fixed compute budgets that explicitly accounts for data composition. Conventional approaches measure training data solely by total tokens, yet the number of examples and their average token length -- what we term \emph{dataset volume} -- play a decisive role in model performance. Our formulation is tuned following established procedures. Experiments on the BRICC dataset \cite{salavati2024reducing} and subsets of the MMLU dataset \cite{hendrycks2021measuringmassivemultitasklanguage}, evaluated under multiple subsampling strategies, reveal that data composition significantly affects token efficiency. These results motivate refined scaling laws for practical LLM fine-tuning in resource-constrained settings.