Practical Benefits of Feature Feedback Under Distribution Shift
Anurag Katakkar, Clay H. Yoo, Weiqin Wang, Zachary Lipton, Divyansh Kaushik
Abstract
In attempts to develop sample-efficient and interpretable algorithms, researcher have explored myriad mechanisms for collecting and exploiting feature feedback, auxiliary annotations provided for training (but not test) instances that highlight salient evidence. Examples include bounding boxes around objects and salient spans in text. Despite its intuitive appeal, feature feedback has not delivered significant gains in practical problems as assessed on iid holdout sets. However, recent works on counterfactually augmented data suggest an alternative benefit of supplemental annotations, beyond interpretability: lessening sensitivity to spurious patterns and consequently delivering gains in out-of-domain evaluations. We speculate that while existing methods for incorporating feature feedback have delivered negligible in-sample performance gains, they may nevertheless provide out-of-domain benefits. Our experiments addressing sentiment analysis, show that feature feedback methods perform significantly better on various natural out-of-domain datasets despite comparable in-domain evaluations. By contrast, performance on natural language inference remains comparable. Finally, we compare those tasks where feature feedback does (and does not) help.- Anthology ID:
- 2022.blackboxnlp-1.29
- Volume:
- Proceedings of the Fifth BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP
- Month:
- December
- Year:
- 2022
- Address:
- Abu Dhabi, United Arab Emirates (Hybrid)
- Editors:
- Jasmijn Bastings, Yonatan Belinkov, Yanai Elazar, Dieuwke Hupkes, Naomi Saphra, Sarah Wiegreffe
- Venue:
- BlackboxNLP
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 346–355
- Language:
- URL:
- https://aclanthology.org/2022.blackboxnlp-1.29
- DOI:
- 10.18653/v1/2022.blackboxnlp-1.29
- Cite (ACL):
- Anurag Katakkar, Clay H. Yoo, Weiqin Wang, Zachary Lipton, and Divyansh Kaushik. 2022. Practical Benefits of Feature Feedback Under Distribution Shift. In Proceedings of the Fifth BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP, pages 346–355, Abu Dhabi, United Arab Emirates (Hybrid). Association for Computational Linguistics.
- Cite (Informal):
- Practical Benefits of Feature Feedback Under Distribution Shift (Katakkar et al., BlackboxNLP 2022)
- PDF:
- https://preview.aclanthology.org/nschneid-patch-4/2022.blackboxnlp-1.29.pdf