TestNUC: Enhancing Test-Time Computing Approaches and Scaling through Neighboring Unlabeled Data Consistency

TL;DR

TestNUC is a novel test-time approach that enhances large language model predictions by leveraging local consistency among neighboring unlabeled data, demonstrating scalable and robust improvements across various tasks.

Contribution

Introduces TestNUC, a linearly scaling method that improves test-time predictions by utilizing neighboring unlabeled data consistency, compatible with existing approaches.

Findings

Consistently outperforms baseline methods across eight datasets.

Scales effectively with increasing unlabeled data.

Robust performance across different embedding models.

Abstract

Test-time computing approaches, which leverage additional computational resources during inference, have been proven effective in enhancing large language model performance. This work introduces a novel, linearly scaling approach, TestNUC, that improves test-time predictions by leveraging the local consistency of neighboring unlabeled data-it classifies an input instance by considering not only the model's prediction on that instance but also on neighboring unlabeled instances. We evaluate TestNUC across eight diverse datasets, spanning intent classification, topic mining, domain discovery, and emotion detection, demonstrating its consistent superiority over baseline methods such as standard prompting and self-consistency. Furthermore, TestNUC can be seamlessly integrated with existing test-time computing approaches, substantially boosting their performance. Our analysis reveals that TestNUC scales effectively with increasing amounts of unlabeled data and performs robustly across different embedding models, making it practical for real-world applications. Our code is available at https://github.com/HenryPengZou/TestNUC.