Generative Active Testing: Efficient LLM Evaluation via Proxy Task Adaptation
This addresses the problem of expensive expert annotation for LLM evaluation in domains like healthcare and biomedicine, offering a scalable solution for cost-effective benchmarking.
The paper tackles the high cost of labeling test samples for benchmarking Large Language Models in specialized domains by proposing Generative Active Testing (GAT), an uncertainty-aware acquisition framework that reduces estimation error by ~40% compared to traditional sampling baselines.
With the widespread adoption of pre-trained Large Language Models (LLM), there exists a high demand for task-specific test sets to benchmark their performance in domains such as healthcare and biomedicine. However, the cost of labeling test samples while developing new benchmarks poses a significant challenge, especially when expert annotators are required. Existing frameworks for active sample selection offer limited support for generative Question Answering tasks, where option dynamics can affect model decision boundaries. In this paper, we present Generative Active Testing (GAT), an uncertainty-aware acquisition framework leveraging LLMs as surrogates for informing the sample selection process. Using a novel Statement Adaptation Module, we modify generative tasks into a pseudo-classification format, enabling the capture of sample-level uncertainties across unlabeled candidates. Our zero-shot acquisition functions reduce estimation error by ~40% compared to traditional sampling baselines, offering a scalable solution for cost-effective model benchmarking.