What’s in Your Head? Emergent Behaviour in Multi-Task Transformer Models
By Mor Geva, Uri Katz, Aviv Ben-Arie, Jonathan Berant. 2021
12 citations recorded by the source corpus.
The primary paradigm for multi-task training in natural language processing is to represent the input with a shared pre-trained language model, and add a small, thin network (head) per task. Given an input, a target head is the head that is selected for outputting the final prediction. In this work, we examine the behaviour of non-target heads, that is, the output of heads when given input that belongs to a different task than the one they were trained for. We find that non-target heads exhibit emergent behaviour, which may either explain the target task, or generalize beyond their original task. For example, in a numerical reasoning task, a span extraction head extracts from the input the arguments to a computation that results in a number generated by a target generative head. In addition, a summarization head that is trained with a target question answering head, outputs query-based summaries when given a question and a context from which the answer is to be extracted. This emergent behaviour suggests that multi-task training leads to non-trivial extrapolation of skills, which can be harnessed for interpretability and generalization.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Large Language Models
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Yi Tay
- Donald Metzler
- Dara Bahri
- Vinh Q. Tran
- Jai Gupta
- Tal Schuster
- Sebastian Ruder
- Mostafa Dehghani
In Computer Vision and Pattern Recognition
- Yi Tay
- Donald Metzler
- Dara Bahri
- Sebastian Ruder
- Jai Gupta
- Vinh Q. Tran
- Vamsi Krishna Aribandi
- Connor Shorten
In Large Language Models
- Yi Tay
- Dara Bahri
- Jai Gupta
- Vamsi Krishna Aribandi
- Donald Metzler
- Taghi M. Khoshgoftaar
- Vinh Q. Tran
- Huaixiu Steven Zheng
Citations in the corpus
- ExT5: Towards Extreme Multi-Task Scaling for Transfer Learning (2021, background): “Recently, Geva et al. (2021) demonstrated and analyzed the behavior transfer in encoder-only Transformers between related jointly-trained tasks such as QA and summarization.”
- Text Data Augmentation for Deep Learning (2021, background): “Geva et al. [142] have also shown utility in sharing a feature extractor base and training separate heads.”