arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2402.14890cs.CLcs.AIcs.LG
收藏

维果茨基距离:基准任务相似性的度量

Vygotsky Distance: Measure for Benchmark Task Similarity

CAIRO, Technical University of Applied Sciences Würzburg-Schweinfurt(维尔茨堡-施韦因富特应用科学大学CAIRO)

Maxim K. Surkov, Ivan P. Yamshchikov

AI总结:

本文提出基于模型相对表现的“维果茨基距离”来度量基准任务相似性,实验表明可减少至少40%的评估任务并验证新任务,从而提升模型泛化潜力。

摘要:

评估在现代自然语言处理中扮演着重要角色。大多数现代自然语言处理基准由任意任务集合组成,这些集合既不能保证模型在测试集之外应用时具有任何泛化潜力,也不试图最小化模型评估所需的资源消耗。本文提出了一种理论工具和一种实用算法来计算基准任务之间的相似性,我们将这种相似性度量称为“维果茨基距离”。该相似性度量的核心思想是,它基于“学生”在给定任务上的相对表现,而不是基于任务本身的属性。如果两个任务在维果茨基距离上彼此接近,那么模型在它们上往往具有相似的相对表现。因此,了解任务之间的维果茨基距离,可以在保持高验证质量的同时显著减少评估任务的数量。在包括GLUE、SuperGLUE、CLUE和RussianSuperGLUE在内的各种基准上的实验表明,绝大多数自然语言处理基准在所含任务方面至少可以减少40%。最重要的是,维果茨基距离还可以用于新任务的验证,从而提高未来自然语言处理模型的泛化潜力。

英文摘要:

Evaluation plays a significant role in modern natural language processing. Most modern NLP benchmarks consist of arbitrary sets of tasks that neither guarantee any generalization potential for the model once applied outside the test set nor try to minimize the resource consumption needed for model evaluation. This paper presents a theoretical instrument and a practical algorithm to calculate similarity between benchmark tasks, we call this similarity measure "Vygotsky distance". The core idea of this similarity measure is that it is based on relative performance of the "students" on a given task, rather that on the properties of the task itself. If two tasks are close to each other in terms of Vygotsky distance the models tend to have similar relative performance on them. Thus knowing Vygotsky distance between tasks one can significantly reduce the number of evaluation tasks while maintaining a high validation quality. Experiments on various benchmarks, including GLUE, SuperGLUE, CLUE, and RussianSuperGLUE, demonstrate that a vast majority of NLP benchmarks could be at least 40% smaller in terms of the tasks included. Most importantly, Vygotsky distance could also be used for the validation of new tasks thus increasing the generalization potential of the future NLP models.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑