Grounding Generative Evaluations of Language Models in Unsupervised Document Corpora

Abstract

Language Models (LMs) continue to advance, improving response quality and coherence. Given Internet-scale training datasets, LMs have likely encountered much of what users may 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 rapidly being outpaced by the size and scope of the models under evaluation. 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 the same LMs to evaluate domain-specific knowledge automatically, using only grounding documents (e.g., a textbook) as input. This generative benchmarking approach corresponds well with human curated questions producing an ensemble Spearman ranking correlation of $0.91$ and a benchmark evaluation Pearson accuracy correlation of $0.74$ (model specific $0.82$). This novel approach 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 three recent documents (two post LM knowledge cutoff), discovering a surprisingly strong performance from Gemma-3 models on open-ended questions. Code is available at \url{https://github.com/mmajurski/generative-lm-benchmarking}