Reward Gaming in Conditional Text Generation
This work highlights a critical problem in aligning text generation models for the natural language generation community, though it is incremental as it builds on existing discussions about reward gaming.
The paper identifies three cases where reward functions in conditional text generation incorrectly assign high rewards to undesirable patterns, such as noise-induced spurious correlation, and shows that these issues can be amplified during reinforcement learning training, potentially leading to misaligned model outputs.
To align conditional text generation model outputs with desired behaviors, there has been an increasing focus on training the model using reinforcement learning (RL) with reward functions learned from human annotations. Under this framework, we identify three common cases where high rewards are incorrectly assigned to undesirable patterns: noise-induced spurious correlation, naturally occurring spurious correlation, and covariate shift. We show that even though learned metrics achieve high performance on the distribution of the data used to train the reward function, the undesirable patterns may be amplified during RL training of the text generation model. While there has been discussion about reward gaming in the RL or safety community, in this discussion piece, we would like to highlight reward gaming in the natural language generation (NLG) community using concrete conditional text generation examples and discuss potential fixes and areas for future work.