Saved in:
Bibliographic Details
Main Author: Waldron, Maja
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.05219
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912862502387712
author Waldron, Maja
author_facet Waldron, Maja
contents One-shot prediction enables rapid adaptation of pretrained foundation models to new tasks using only one labeled example, but lacks principled uncertainty quantification. While conformal prediction provides finite-sample coverage guarantees, standard split conformal methods are inefficient in the one-shot setting due to data splitting and reliance on a single predictor. We propose Conformal Aggregation of One-Shot Predictors (CAOS), a conformal framework that adaptively aggregates multiple one-shot predictors and uses a leave-one-out calibration scheme to fully exploit scarce labeled data. Despite violating classical exchangeability assumptions, we prove that CAOS achieves valid marginal coverage using a monotonicity-based argument. Experiments on one-shot facial landmarking and RAFT text classification tasks show that CAOS produces substantially smaller prediction sets than split conformal baselines while maintaining reliable coverage.
format Preprint
id arxiv_https___arxiv_org_abs_2601_05219
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle CAOS: Conformal Aggregation of One-Shot Predictors
Waldron, Maja
Machine Learning
Artificial Intelligence
One-shot prediction enables rapid adaptation of pretrained foundation models to new tasks using only one labeled example, but lacks principled uncertainty quantification. While conformal prediction provides finite-sample coverage guarantees, standard split conformal methods are inefficient in the one-shot setting due to data splitting and reliance on a single predictor. We propose Conformal Aggregation of One-Shot Predictors (CAOS), a conformal framework that adaptively aggregates multiple one-shot predictors and uses a leave-one-out calibration scheme to fully exploit scarce labeled data. Despite violating classical exchangeability assumptions, we prove that CAOS achieves valid marginal coverage using a monotonicity-based argument. Experiments on one-shot facial landmarking and RAFT text classification tasks show that CAOS produces substantially smaller prediction sets than split conformal baselines while maintaining reliable coverage.
title CAOS: Conformal Aggregation of One-Shot Predictors
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2601.05219