Papers by Simon D. Angus
Models Teaching Models: Improving Model Accuracy with Slingshot Learning (2023.eacl-main)
Copied to clipboard
| Challenge: | Existing methods to label datasets are expensive and require human labor . a semi-supervised method that augments a small dataset with labels reduces the cost of using simpler methods . |
| Approach: | They propose a semi-supervised method to augment a human-labeled dataset with labels from a teacher model to slingshot the performance of a student model. |
| Outcome: | The proposed method reduces the accuracy trade-off required to use simpler methods without disrupting their benefits. |