RepMatch: Quantifying Cross-Instance Similarities in Representation Space

Mohammad Modarres, Sina Abbasi, Mohammad Taher Pilehvar


Abstract
Advances in dataset analysis techniques have enabled more sophisticated approaches to analyzing and characterizing training data instances, often categorizing data based on attributes such as “difficulty”. In this work, we introduce RepMatch, a novel method that characterizes data through the lens of similarity.RepMatch quantifies the similarity between subsets of training instances by comparing the knowledge encoded in models trained on them, overcoming the limitations of existing analysis methods that focus solely on individual instances and are restricted to within-dataset analysis.Our framework allows for a broader evaluation, enabling similarity comparisons across arbitrary subsets of instances, supporting both dataset-to-dataset and instance-to-dataset analyses. We validate the effectiveness of RepMatch across multiple NLP tasks, datasets, and models. Through extensive experimentation, we demonstrate that RepMatch can effectively compare datasets, identify more representative subsets of a dataset (that lead to better performance than randomly selected subsets of equivalent size), and uncover heuristics underlying the construction of some challenge datasets.
Anthology ID:
2024.emnlp-main.825
Volume:
Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing
Month:
November
Year:
2024
Address:
Miami, Florida, USA
Editors:
Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
Venue:
EMNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
14858–14869
Language:
URL:
https://aclanthology.org/2024.emnlp-main.825
DOI:
Bibkey:
Cite (ACL):
Mohammad Modarres, Sina Abbasi, and Mohammad Taher Pilehvar. 2024. RepMatch: Quantifying Cross-Instance Similarities in Representation Space. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 14858–14869, Miami, Florida, USA. Association for Computational Linguistics.
Cite (Informal):
RepMatch: Quantifying Cross-Instance Similarities in Representation Space (Modarres et al., EMNLP 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.emnlp-main.825.pdf