No abstract

Large pretrained models have seen enormous success in extractive summarization tasks. We investigate, here, the influence of pretraining on a BERT-based extractive summarization system for scientific documents. We derive performance improvements using an intermediate pretraining step that leverages existing summarization datasets and report state-of-theart results on a recently released scientific summarization dataset, SCITLDR. We systematically analyze the intermediate pretraining step by varying the size and domain of the pretraining corpus, changing the length of the input sequence in the target task and varying target tasks. We also investigate how intermediate pretraining interacts with contextualized word embeddings trained on different domains.

State abstraction is necessary for better task transfer in complex reinforcement learning environments. Inspired by the benefit of state abstraction in MAXQ and building upon hybrid planner-RL architectures, we propose RePReL, a hierarchical framework that leverages a relational planner to provide useful state abstractions. Our experiments demonstrate that the abstractions enable faster learning and efficient transfer across tasks. More importantly, our framework enables the application of standard RL approaches for learning in structured domains. The benefit of using the state abstractions is critical in relational settings, where the number and/or types of objects are not fixed apriori. Our experiments clearly show that RePReL framework not only achieves better performance and efficient learning on the task at hand but also demonstrates better generalization to unseen tasks.

No abstract