Abstract
Originally proposed as a method for knowledge transfer from one model to another, some recent studies have suggested that knowledge distillation (KD) is in fact a form of regularization. Perhaps the strongest argument of all for this new perspective comes from its apparent similarities with label smoothing (LS). Here we re-examine this stated equivalence between the two methods by comparing the predictive confidences of the models they train. Experiments on four text classification tasks involving models of different sizes show that: (a) In most settings, KD and LS drive model confidence in completely opposite directions, and (b) In KD, the student inherits not only its knowledge but also its confidence from the teacher, reinforcing the classical knowledge transfer view.- Anthology ID:
- 2023.emnlp-main.271
- Volume:
- Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing
- Month:
- December
- Year:
- 2023
- Address:
- Singapore
- Editors:
- Houda Bouamor, Juan Pino, Kalika Bali
- Venue:
- EMNLP
- SIG:
- Publisher:
- Association for Computational Linguistics
- Note:
- Pages:
- 4469–4477
- Language:
- URL:
- https://aclanthology.org/2023.emnlp-main.271
- DOI:
- 10.18653/v1/2023.emnlp-main.271
- Cite (ACL):
- Md Sultan. 2023. Knowledge Distillation ≈ Label Smoothing: Fact or Fallacy?. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 4469–4477, Singapore. Association for Computational Linguistics.
- Cite (Informal):
- Knowledge Distillation ≈ Label Smoothing: Fact or Fallacy? (Sultan, EMNLP 2023)
- PDF:
- https://preview.aclanthology.org/emnlp-22-attachments/2023.emnlp-main.271.pdf