Human-centric Dialog Training via Offline Reinforcement Learning

by   Natasha Jaques, et al.

How can we train a dialog model to produce better conversations by learning from human feedback, without the risk of humans teaching it harmful chat behaviors? We start by hosting models online, and gather human feedback from real-time, open-ended conversations, which we then use to train and improve the models using offline reinforcement learning (RL). We identify implicit conversational cues including language similarity, elicitation of laughter, sentiment, and more, which indicate positive human feedback, and embed these in multiple reward functions. A well-known challenge is that learning an RL policy in an offline setting usually fails due to the lack of ability to explore and the tendency to make over-optimistic estimates of future reward. These problems become even harder when using RL for language models, which can easily have a 20,000 action vocabulary and many possible reward functions. We solve the challenge by developing a novel class of offline RL algorithms. These algorithms use KL-control to penalize divergence from a pre-trained prior language model, and use a new strategy to make the algorithm pessimistic, instead of optimistic, in the face of uncertainty. We test the resulting dialog model with ratings from 80 users in an open-domain setting and find it achieves significant improvements over existing deep offline RL approaches. The novel offline RL method is viable for improving any existing generative dialog model using a static dataset of human feedback.


page 6

page 19


Way Off-Policy Batch Deep Reinforcement Learning of Implicit Human Preferences in Dialog

Most deep reinforcement learning (RL) systems are not able to learn effe...

NaRLE: Natural Language Models using Reinforcement Learning with Emotion Feedback

Current research in dialogue systems is focused on conversational assist...

Improving Zero-shot Generalization in Offline Reinforcement Learning using Generalized Similarity Functions

Reinforcement learning (RL) agents are widely used for solving complex s...

Reinforced Self-Training (ReST) for Language Modeling

Reinforcement learning from human feedback (RLHF) can improve the qualit...

Improving Language Models with Advantage-based Offline Policy Gradients

Improving language model generations according to some user-defined qual...

Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback

We apply preference modeling and reinforcement learning from human feedb...

Integrating Pretrained Language Model for Dialogue Policy Learning

Reinforcement Learning (RL) has been witnessed its potential for trainin...

Please sign up or login with your details

Forgot password? Click here to reset