Evaluating Language Models as Synthetic Data Generators

Given the increasing use of synthetic data in language model (LM) post-training, an LM's ability to generate high-quality data has become nearly as crucial as its ability to solve problems directly. While prior works have focused on developing effective data generation methods, they lack systematic comparison of different LMs as data generators in a unified setting. To address this gap, we propose AGORABENCH, a benchmark that provides standardized settings and metrics to evaluate LMs' data generation abilities. Through synthesizing 1.26 million training instances using 6 LMs and training 99 student models, we uncover key insights about LMs' data generation capabilities. First, we observe that LMs exhibit distinct strengths. For instance, GPT-4o excels at generating new problems, while Claude-3.5-Sonnet performs better at enhancing existing ones. Furthermore, our analysis reveals that an LM's data generation ability doesn't necessarily correlate with its problem-solving ability. Instead, multiple intrinsic features of data quality-including response quality, perplexity, and instruction difficulty-collectively serve as better indicators. Finally, we demonstrate that strategic choices in output format and costconscious model selection significantly impact data generation effectiveness. Our code, checkpoints, and data are all publicly available at https://github.com/neulab/data-agora . What is the most effective data generation method when using Llama-3? GPT-3 Instruct GPT Chat GPT GPT-4 Llama-3 70B Quality Enhancement Data Generation Methods Data Generator Data Generator Data Generation Methods Instance Generation Response Generation Is there a big difference between using GPT-4o and GPT-4o-mini as data generators? Self Instruct Alpaca Conventional Setting AgoraBench (Ours) Wizard LM Orca Magpie Llama-3.1 8,70,405B GPT-4o Comparable Not Comparable GPT-4o mini