How Many Data Samples is an Additional Instruction Worth?
This addresses the challenge for non-expert users who cannot easily create large datasets or modify model architectures, offering a practical alternative to enhance task performance.
The paper tackles the problem of improving instruction-tuned NLP models for non-expert users by augmenting tasks with additional instructions, finding that this approach significantly boosts performance by up to 35%, with an additional instruction being equivalent to about 200 data samples on average.
Recently introduced instruction-paradigm empowers non-expert users to leverage NLP resources by defining a new task in natural language. Instruction-tuned models have significantly outperformed multitask learning models (without instruction); however they are far from state-of-the-art task-specific models. Conventional approaches to improve model performance via creating datasets with large number of task instances or architectural changes in the model may not be feasible for non-expert users. However, they can write alternate instructions to represent an instruction task. Is Instruction-augmentation helpful? We augment a subset of tasks in the expanded version of NATURAL INSTRUCTIONS with additional instructions and find that it significantly improves model performance (up to 35%), especially in the low-data regime. Our results indicate that an additional instruction can be equivalent to ~200 data samples on average across tasks.