From Quantity to Quality: Boosting LLM Performance with Self-Guided Data Selection for Instruction Tuning
This addresses the challenge of manual curation and high costs in instruction tuning for LLM developers, representing an incremental efficiency improvement.
The paper tackles the problem of balancing instruction data quality and quantity for LLMs by introducing a self-guided method using an Instruction-Following Difficulty metric to select high-quality samples, achieving improved results with only 10% of the original data.
In the realm of Large Language Models (LLMs), the balance between instruction data quality and quantity is a focal point. Recognizing this, we introduce a self-guided methodology for LLMs to autonomously discern and select cherry samples from open-source datasets, effectively minimizing manual curation and potential cost for instruction tuning an LLM. Our key innovation, the Instruction-Following Difficulty (IFD) metric, emerges as a pivotal metric to identify discrepancies between a model's expected responses and its intrinsic generation capability. Through the application of IFD, cherry samples can be pinpointed, leading to a marked uptick in model training efficiency. Empirical validations on datasets like Alpaca and WizardLM underpin our findings; with a mere $10\%$ of original data input, our strategy showcases improved results. This synthesis of self-guided cherry-picking and the IFD metric signifies a transformative leap in the instruction tuning of LLMs, promising both efficiency and resource-conscious advancements. Codes, data, and models are available: https://github.com/tianyi-lab/Cherry_LLM