Grounding Synthetic Data Evaluations of Language Models in Unsupervised Document Corpora

Language Models (LMs) continue to advance, improving response quality and coherence. Given Internet-scale training datasets, LMs have likely encountered much of what users might ask them to generate in some form during their training. A plethora of evaluation benchmarks have been constructed to assess model quality, response appropriateness, and reasoning capabilities. However, the human effort required for benchmark construction is limited and being rapidly outpaced by the size and scope of the models under evaluation. Additionally, having humans build a benchmark for every possible domain of interest is impractical. Therefore, we propose a methodology for automating the construction of fact-based synthetic data model evaluations grounded in document populations. This work leverages those very same LMs to evaluate domain-specific knowledge automatically, using only grounding documents (e.g., a textbook) as input. This synthetic data benchmarking approach corresponds well with human curated questions with a Spearman ranking correlation of 0.96 and a benchmark evaluation Pearson accuracy correlation of 0.79. This novel tool supports generating both multiple choice and open-ended synthetic data questions to gain diagnostic insight of LM capability. We apply this methodology to evaluate model performance on a recent relevant arXiv preprint, discovering a surprisingly strong performance from Gemma3 models.