Abstract
We study few-shot debugging of transformer based natural language understanding models, using recently popularized test suites to not just diagnose but correct a problem. Given a few debugging examples of a certain phenomenon, and a held-out test set of the same phenomenon, we aim to maximize accuracy on the phenomenon at a minimal cost of accuracy on the original test set. We examine several methods that are faster than full epoch retraining. We introduce a new fast method, which samples a few in-danger examples from the original training set. Compared to fast methods using parameter distance constraints or Kullback-Leibler divergence, we achieve superior original accuracy for comparable debugging accuracy.- Anthology ID:
- 2022.deelio-1.8
- Volume:
- Proceedings of Deep Learning Inside Out (DeeLIO 2022): The 3rd Workshop on Knowledge Extraction and Integration for Deep Learning Architectures
- Month:
- May
- Year:
- 2022
- Address:
- Dublin, Ireland and Online
- Venue:
- DeeLIO
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 79–86
- Language:
- URL:
- https://aclanthology.org/2022.deelio-1.8
- DOI:
- 10.18653/v1/2022.deelio-1.8
- Cite (ACL):
- Christopher Malon, Kai Li, and Erik Kruus. 2022. Fast Few-shot Debugging for NLU Test Suites. In Proceedings of Deep Learning Inside Out (DeeLIO 2022): The 3rd Workshop on Knowledge Extraction and Integration for Deep Learning Architectures, pages 79–86, Dublin, Ireland and Online. Association for Computational Linguistics.
- Cite (Informal):
- Fast Few-shot Debugging for NLU Test Suites (Malon et al., DeeLIO 2022)
- PDF:
- https://preview.aclanthology.org/paclic-22-ingestion/2022.deelio-1.8.pdf
- Code
- necla-ml/debug-test-suites
- Data
- GLUE, MultiNLI, SST