From Real to Synthetic: Synthesizing Millions of Diversified and Complicated User Instructions with Attributed Grounding
This addresses the need for large-scale, high-quality instruction data to improve LLM alignment, representing an incremental advance in synthetic data generation methods.
The paper tackles the problem of generating diverse and complex instruction data for aligning large language models by synthesizing 1 million instructions using attributed grounding, resulting in models trained on this dataset achieving leading performance on several benchmarks with scalable improvements.
The pursuit of diverse, complex, and large-scale instruction data is crucial for automatically aligning large language models (LLMs). While there are methods capable of generating synthetic instructions at scale, they either suffer from limited grounding sources, leading to a narrow distribution, or rely on trivial extensions that fail to produce meaningful trajectories in terms of complexity. In contrast, instructions that benefit efficient alignment are typically crafted with cognitive insights and grounded in real-world use cases. In this paper, we synthesize such instructions using attributed grounding, which involves 1) a top-down attribution process that grounds a selective set of real instructions to situated users, and 2) a bottom-up synthesis process that leverages web documents to first generate a situation, then a meaningful instruction. This framework allows us to harvest diverse and complex instructions at scale, utilizing the vast range of web documents. Specifically, we construct a dataset of 1 million instructions, called SynthQuestions, and demonstrate that models trained on it achieve leading performance on several common benchmarks, with improvements that continually scale with more web corpora. Data, models and codes will be available at https://github.com/Ignoramus0817/SynthQuestions.