LGAIMEMay 8

Query-efficient model evaluation using cached responses

arXiv:2605.0709653.32 citations
AI Analysis

For practitioners needing to evaluate models on expensive benchmarks, this method reduces query costs by exploiting cached responses from previously evaluated models.

This paper introduces a method using Data Kernel Perspective Space (DKPS) to leverage cached model responses for query-efficient evaluation of new models on benchmarks, achieving the same mean absolute error as baselines with substantially fewer queries.

Evaluating a new model on an existing benchmark is often necessary to understand its behavior before deployment. For modern evaluation frameworks, generating and evaluating a response for all queries can be prohibitively expensive. In practice, responses from previously-evaluated models are often cached -- creating a potential opportunity to use this additional information to decrease the number of queries required to accurately evaluate a new model. In this paper, we introduce an approach for predicting benchmark performance that leverages cached model responses based on the Data Kernel Perspective Space (DKPS), a method for quantifying the relationship between models in the black-box setting. Theoretically, we show that DKPS-based methods are query-efficient under certain conditions. Empirically, we demonstrate that DKPS-based methods achieve the same mean absolute error as baselines with a substantially decreased query budget. We conclude by proposing an offline method for selecting a set of queries that maximizes the goodness-of-fit on reference models, improving prediction accuracy over random query selection.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes