# LOTTERY: Learning from Reference-Only Samples in Two-Sample Testing under Size Asymmetry

Source: [LOTTERY: Learning from Reference-Only Samples in Two-Sample Testing under Size Asymmetry](<https://www.sota2.com/research/paper/6a279e1c90966ade919bf721>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-06-10T05:01:35.776Z

Authors: Xunye Tian, Zhijian Zhou, Liuhua Peng, Feng Liu

Published: 2026-06-07T05:49:42.000Z

## Abstract

Data-adaptive two-sample testing assesses if two samples come from the same distribution, using a discrepancy learned from the data (e.g., via kernel-based feature representations). Such methods typically rely on data splitting to decouple learning from testing and control type I error. However, this paradigm is ill-suited to few-shot settings with severe sample-size imbalance: abundant reference samples are available, while only a handful of query samples arrive. In this paper, we show how this imbalance can be leveraged constructively. Using abundant reference data, we learn reference-dependent representations that summarize salient structure of the reference distribution and provide informative signals for detecting departures. We incorporate a collection of representation families that capture both global and local structure, and adaptively weight them using only reference samples via an uncertainty-guided principle. Theoretically, we establish permutation-based type I error control and show consistency of the aggregated test: as the sample sizes grow, the test power converges to one whenever the representation set contains at least one consistent representation. Empirically, our aggregation achieves strong performance across a range of benchmarks while retaining type I error control.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Two-Sample Testing on CIFAR10-RES18 (test)](<https://www.sota2.com/research/sota/two-sample-testing-on-cifar10-res18-test.md>): Two-sample testing; CIFAR10-RES18 (test); Test Power: 100.

- [Two-Sample Testing on CIFAR10 WRN28](<https://www.sota2.com/research/sota/two-sample-testing-on-cifar10-wrn28.md>): Two-sample testing; CIFAR10 WRN28; Test Power: 66.9.

- [Two-Sample Testing on BLOB (test)](<https://www.sota2.com/research/sota/two-sample-testing-on-blob-test.md>): Two-sample testing; BLOB (test); Test Power: 72.5.

- [Two-Sample Testing on BLOB](<https://www.sota2.com/research/sota/two-sample-testing-on-blob.md>): Two-sample testing; Blob; Test Power: 1.

- [Two-Sample Testing on CIFAR10-WRN8](<https://www.sota2.com/research/sota/two-sample-testing-on-cifar10-wrn8.md>): Two-sample testing; CIFAR10-WRN8; Test Power: 95.8.

- [Two-sample testing on HIGGS](<https://www.sota2.com/research/sota/two-sample-testing-on-higgs.md>): Two-sample testing; higgs; Test Power: 100.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2606.08460>)

[PDF](<https://arxiv.org/pdf/2606.08460.pdf>)
