Efficient (Soft) Q-Learning for Text Generation with Limited Good Data
This addresses the problem of flexible text generation for applications lacking direct supervision, such as adversarial attacks or prompt control, with an incremental improvement over prior RL methods.
The paper tackles the inefficiency and instability of existing reinforcement learning methods for text generation by introducing a new soft Q-learning formulation that combines on- and off-policy updates to learn effectively from sparse rewards. Experiments show it consistently outperforms both task-specialized algorithms and previous RL methods across tasks like learning from noisy examples, adversarial attacks, and prompt generation.
Maximum likelihood estimation (MLE) is the predominant algorithm for training text generation models. This paradigm relies on direct supervision examples, which is not applicable to many emerging applications, such as generating adversarial attacks or generating prompts to control language models. Reinforcement learning (RL) on the other hand offers a more flexible solution by allowing users to plug in arbitrary task metrics as reward. Yet previous RL algorithms for text generation, such as policy gradient (on-policy RL) and Q-learning (off-policy RL), are often notoriously inefficient or unstable to train due to the large sequence space and the sparse reward received only at the end of sequences. In this paper, we introduce a new RL formulation for text generation from the soft Q-learning (SQL) perspective. It enables us to draw from the latest RL advances, such as path consistency learning, to combine the best of on-/off-policy updates, and learn effectively from sparse reward. We apply the approach to a wide range of novel text generation tasks, including learning from noisy/negative examples, adversarial attacks, and prompt generation. Experiments show our approach consistently outperforms both task-specialized algorithms and the previous RL methods.