From Symbolic Tasks to Code Generation: Diversification Yields Better Task Performers

TL;DR

This paper demonstrates that diversifying instruction-tuning data across a broad semantic space enhances large language models' ability to generalize and follow instructions, especially in code generation tasks.

Contribution

It introduces a theoretical framework using Markov algorithms to analyze instruction tuning and shows that diversity in training tasks improves model robustness and generalization.

Findings

Diverse instruction sets improve generalization in synthetic experiments.

Increased task diversity enhances code generation performance.

Robustness to distribution shifts increases with instruction set diversity.

Abstract

Instruction tuning -- tuning large language models on instruction-output pairs -- is a promising technique for making models better adapted to the real world. Yet, the key factors driving the model's capability to understand and follow instructions not seen during training remain under-explored. Our investigation begins with a series of synthetic experiments within the theoretical framework of a Turing-complete algorithm called Markov algorithm, which allows fine-grained control over the instruction-tuning data. Generalization and robustness with respect to the training distribution emerge once a diverse enough set of tasks is provided, even though very few examples are provided for each task. We extend these initial results to a real-world application scenario of code generation and find that a more diverse instruction set, extending beyond code-related tasks, improves the performance of code generation. Our observations suggest that a more diverse semantic space for instruction-tuning sets greatly improves the model's ability to follow instructions and perform tasks.