Hi,
Thank you for your continuous efforts in maintaining and updating this excellent repository.
I would like to kindly recommend our recent work, "TestNUC: Enhancing Test-Time Computing Approaches and Scaling through Neighboring Unlabeled Data Consistency", which has been accepted at ACL 2025. In this paper, we introduce 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. TestNUC scales effectively with increasing amounts of unlabeled data and performs robustly across different embedding models, making it practical for real-world applications.
TestNUC can also be seamlessly integrated with existing test-time computing approaches, substantially boosting their performance.
You can find our paper here: https://arxiv.org/abs/2502.19163
Code: https://github.com/HenryPengZou/TestNUC
We believe it could be a relevant and valuable addition to your repository. Thank you very much for your time and consideration.
Best regards,
Hi,
Thank you for your continuous efforts in maintaining and updating this excellent repository.
I would like to kindly recommend our recent work, "TestNUC: Enhancing Test-Time Computing Approaches and Scaling through Neighboring Unlabeled Data Consistency", which has been accepted at ACL 2025. In this paper, we introduce 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. TestNUC scales effectively with increasing amounts of unlabeled data and performs robustly across different embedding models, making it practical for real-world applications.
TestNUC can also be seamlessly integrated with existing test-time computing approaches, substantially boosting their performance.
You can find our paper here: https://arxiv.org/abs/2502.19163
Code: https://github.com/HenryPengZou/TestNUC
We believe it could be a relevant and valuable addition to your repository. Thank you very much for your time and consideration.
Best regards,