X-SHOT: A Single System to Handle Frequent, Few-shot and Zero-shot Labels in Classification
Keywords: Natural Language Processing, Few-shot Learning, Zero-shot Learning
TL;DR: X-shot: a system capable of handling freq-shot, few-shot, and zero-shot problems simultaneously without constraints.
Abstract: In recent years, few-shot and zero-shot learning, which focus on labels with limited annotated instances, have garnered significant attention. Traditional approaches often treat freq-shot (labels with numerous instances), few-shot, and zero-shot learning as distinct challenges, optimizing systems for just one of these scenarios. Yet, in real-world settings, label occurrences vary greatly. Some labels might appear thousands of times, while others might only appear sporadically or not at all. Ideally, a system should accommodate any label, regardless of its training frequency. Notably, while few-shot systems often falter on zero-shot tasks, zero-shot systems don't leverage available annotations when certain downstream labels possess them. For practical deployment, it's crucial that a system can adapt to any label occurrence. We introduce a novel classification challenge: $X$-shot, reflecting a real-world context where freq-shot, few-shot, and zero-shot labels emerge without predefined limits. Here, $X$ can span from 0 to positive infinity. The crux of $X$-shot centers on open-domain generalization and devising a system versatile enough to manage various label scenarios. Our solution leverages Instruction Learning, bolstered by data autonomously generated by pre-trained Language Models (PLMs). Our unified system, $X$-shot, surpasses preceding state-of-the-art techniques on three benchmark datasets across diverse domains in both single-label and multi-label classifications. This is the first work addressing $X$-shot learning, where $X$ remains variable.
Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 6603
Loading