Search:
Match:
1 results

Analysis

This paper proposes a novel method to characterize transfer learning effects by analyzing multi-task learning curves. Instead of focusing on model updates, the authors perturb the dataset size to understand how performance changes. This approach offers a potentially more fundamental understanding of transfer, especially in the context of foundation models. The use of learning curves allows for a quantitative assessment of transfer effects, including pairwise and contextual transfer.
Reference

Learning curves can better capture the effects of multi-task learning and their multi-task extensions can delineate pairwise and contextual transfer effects in foundation models.