School GPT3 for OpenQA We work on engineering the GPT-3 input to improve its performance on the OpenQA task in the in-context learning setting. Task-Oriented Dialog Systems With the goal of reducing the amount of natural language encoded at each turn, we investigated the effect of dialogue history in multilingual task-oriented dialogue systems. Graph Neural Networks In this blog, We will introduce to you the principle and implementation of commonly used tricks in GNN and analyze their effectiveness with experimental results.
