Leveraging Unstructured Text Data for Federated Instruction Tuning of Large Language Models

TL;DR

This paper introduces FedIT-U2S, a framework that automatically converts unstructured text data into structured instruction-response pairs for federated instruction tuning of large language models, broadening its practical applicability.

Contribution

FedIT-U2S is a novel framework that enables federated instruction tuning using unstructured text data by automatic data generation and retrieval-based example selection.

Findings

Consistent improvement over base LLM across three domains.

Effective automatic transformation of unstructured data into instruction-response pairs.

Enhanced flexibility in federated instruction tuning scenarios.

Abstract

Federated instruction tuning enables multiple clients to collaboratively fine-tune a shared large language model (LLM) that can follow humans' instructions without directly sharing raw data. However, existing literature impractically requires that all the clients readily hold instruction-tuning data (i.e., structured instruction-response pairs), which necessitates massive human annotations since clients' data is usually unstructured text instead. Addressing this, we propose a novel and flexible framework FedIT-U2S, which can automatically transform unstructured corpus into structured data for federated instruction tuning. FedIT-U2S consists two key steps: (1) few-shot instruction-tuning data generation, where each unstructured data piece together with several examples is combined to prompt an LLM in generating an instruction-response pair. To further enhance the flexibility, a retrieval-based example selection technique is proposed, where the examples are automatically selected based on the relatedness between the client's data piece and example pool, bypassing the need of determining examples in advance. (2) A typical federated instruction tuning process based on the generated data. Overall, FedIT-U2S can be applied to diverse scenarios as long as the client holds valuable text corpus, broadening the application scope of federated instruction tuning. We conduct a series of experiments on three domains (medicine, knowledge, and math), showing that our proposed FedIT-U2S can consistently and significantly brings improvement over the base LLM.