Less is More: Parameter-Efficient Selection of Intermediate Tasks for Transfer Learning
This work addresses the inefficiency in task selection for transfer learning in NLP, enabling scalable applications with large source pools.
The paper tackles the problem of selecting intermediate tasks for transfer learning efficiently, introducing Embedding Space Maps (ESMs) to approximate fine-tuning effects, and achieves a 10x reduction in execution time and 278x in disk space usage while maintaining high selection performance with an average regret@5 score of 2.95.
Intermediate task transfer learning can greatly improve model performance. If, for example, one has little training data for emotion detection, first fine-tuning a language model on a sentiment classification dataset may improve performance strongly. But which task to choose for transfer learning? Prior methods producing useful task rankings are infeasible for large source pools, as they require forward passes through all source language models. We overcome this by introducing Embedding Space Maps (ESMs), light-weight neural networks that approximate the effect of fine-tuning a language model. We conduct the largest study on NLP task transferability and task selection with 12k source-target pairs. We find that applying ESMs on a prior method reduces execution time and disk space usage by factors of 10 and 278, respectively, while retaining high selection performance (avg. regret@5 score of 2.95).