Research

Paper

TESTING February 24, 2026

DANCE: Doubly Adaptive Neighborhood Conformal Estimation

Authors

Brandon R. Feng, Brian J. Reich, Daniel Beaglehole, Xihaier Luo, David Keetae Park, Shinjae Yoo, Zhechao Huang, Xueyu Mao, Olcay Boz, Jungeum Kim

Abstract

The recent developments of complex deep learning models have led to unprecedented ability to accurately predict across multiple data representation types. Conformal prediction for uncertainty quantification of these models has risen in popularity, providing adaptive, statistically-valid prediction sets. For classification tasks, conformal methods have typically focused on utilizing logit scores. For pre-trained models, however, this can result in inefficient, overly conservative set sizes when not calibrated towards the target task. We propose DANCE, a doubly locally adaptive nearest-neighbor based conformal algorithm combining two novel nonconformity scores directly using the data's embedded representation. DANCE first fits a task-adaptive kernel regression model from the embedding layer before using the learned kernel space to produce the final prediction sets for uncertainty quantification. We test against state-of-the-art local, task-adapted and zero-shot conformal baselines, demonstrating DANCE's superior blend of set size efficiency and robustness across various datasets.

Metadata

arXiv ID: 2602.20652
Provider: ARXIV
Primary Category: stat.ML
Published: 2026-02-24
Fetched: 2026-02-25 06:05

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2602.20652v1</id>\n    <title>DANCE: Doubly Adaptive Neighborhood Conformal Estimation</title>\n    <updated>2026-02-24T07:54:53Z</updated>\n    <link href='https://arxiv.org/abs/2602.20652v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2602.20652v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>The recent developments of complex deep learning models have led to unprecedented ability to accurately predict across multiple data representation types. Conformal prediction for uncertainty quantification of these models has risen in popularity, providing adaptive, statistically-valid prediction sets. For classification tasks, conformal methods have typically focused on utilizing logit scores. For pre-trained models, however, this can result in inefficient, overly conservative set sizes when not calibrated towards the target task. We propose DANCE, a doubly locally adaptive nearest-neighbor based conformal algorithm combining two novel nonconformity scores directly using the data's embedded representation. DANCE first fits a task-adaptive kernel regression model from the embedding layer before using the learned kernel space to produce the final prediction sets for uncertainty quantification. We test against state-of-the-art local, task-adapted and zero-shot conformal baselines, demonstrating DANCE's superior blend of set size efficiency and robustness across various datasets.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='stat.ML'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <published>2026-02-24T07:54:53Z</published>\n    <arxiv:primary_category term='stat.ML'/>\n    <author>\n      <name>Brandon R. Feng</name>\n    </author>\n    <author>\n      <name>Brian J. Reich</name>\n    </author>\n    <author>\n      <name>Daniel Beaglehole</name>\n    </author>\n    <author>\n      <name>Xihaier Luo</name>\n    </author>\n    <author>\n      <name>David Keetae Park</name>\n    </author>\n    <author>\n      <name>Shinjae Yoo</name>\n    </author>\n    <author>\n      <name>Zhechao Huang</name>\n    </author>\n    <author>\n      <name>Xueyu Mao</name>\n    </author>\n    <author>\n      <name>Olcay Boz</name>\n    </author>\n    <author>\n      <name>Jungeum Kim</name>\n    </author>\n  </entry>"
}