Teacher Forcing Recovers Reward Functions for Text GenerationDownload PDF

Published: 31 Oct 2022, 18:00, Last Modified: 05 Jan 2023, 21:49NeurIPS 2022 AcceptReaders: Everyone
Keywords: Text Generation, Natural Language Processing, Reinforcement Learning
TL;DR: We derive a reward function for text generation via the lens of inverse reinforcement learning.
Abstract: Reinforcement learning (RL) has been widely used in text generation to alleviate the exposure bias issue or to utilize non-parallel datasets. The reward function plays an important role in making RL training successful. However, previous reward functions are typically task-specific and sparse, restricting the use of RL. In our work, we propose a task-agnostic approach that derives a step-wise reward function directly from a model trained with teacher forcing. We additionally propose a simple modification to stabilize the RL training on non-parallel datasets with our induced reward function. Empirical results show that our method outperforms self-training and reward regression methods on several text generation tasks, confirming the effectiveness of our reward function.
Supplementary Material: pdf
24 Replies