Tight Bounds for Answering Adaptively Chosen Concentrated Queries

The work by Bassily and Freund introduced the concentrated queries framework in 2016, aiming to address the complexities of adaptive data analysis when sample correlations exist. Their recent findings, published in November 2025, demonstrate that the current framework imposes a significant limitation on the number of queries that can be effectively utilized in an adaptive setting, capping it at O(n) for a sample size of n, in stark contrast to the O(n^2) possible in independent scenarios. This inherent utility gap underscores the challenges faced by analysts and suggests that further refinements to the concentrated queries framework are necessary to enhance its applicability and effectiveness in real-world data analysis tasks.