
The author introduces SMART, a data mixture strategy for instruction tuning, utilizing submodular functions to assign importance scores to tasks and select non-redundant samples. The approach outperforms traditional methods and highlights the importance of task composition in instruction tuning.


coremsg

smart-submodular-data-mixture-strategy-for-instruction-tuning


SMART: Submodular Data Mixture Strategy for Instruction Tuning


title_rewrite


SMARTは、指示調整のためのサブモジュラーデータ混合戦略を導入し、従来の手法を大幅に上回ることを実証します。



Proposing a dual instruction tuning strategy to enhance mathematical reasoning with large language models.


dual-instruction-tuning-strategy-for-mathematical-reasoning-with-large-language-models


Dual Instruction Tuning Strategy for Mathematical Reasoning with Large Language Models
