@inproceedings{ye-etal-2021-towards,
    title = "Towards More Fine-grained and Reliable {NLP} Performance Prediction",
    author = "Ye, Zihuiwen  and
      Liu, Pengfei  and
      Fu, Jinlan  and
      Neubig, Graham",
    editor = "Merlo, Paola  and
      Tiedemann, Jorg  and
      Tsarfaty, Reut",
    booktitle = "Proceedings of the 16th Conference of the European Chapter of the Association for Computational Linguistics: Main Volume",
    month = apr,
    year = "2021",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/fix-sig-urls/2021.eacl-main.324/",
    doi = "10.18653/v1/2021.eacl-main.324",
    pages = "3703--3714",
    abstract = "Performance prediction, the task of estimating a system{'}s performance without performing experiments, allows us to reduce the experimental burden caused by the combinatorial explosion of different datasets, languages, tasks, and models. In this paper, we make two contributions to improving performance prediction for NLP tasks. First, we examine performance predictors not only for holistic measures of accuracy like F1 or BLEU, but also \textit{fine-grained} performance measures such as accuracy over individual classes of examples. Second, we propose methods to understand the \textit{reliability} of a performance prediction model from two angles: confidence intervals and calibration. We perform an analysis of four types of NLP tasks, and both demonstrate the feasibility of fine-grained performance prediction and the necessity to perform reliability analysis for performance prediction methods in the future."
}