Rethinking Reinforcement Learning for Recommendation: A Prompt Perspective

TL;DR

This paper introduces Prompt-Based Reinforcement Learning (PRL), a novel offline training paradigm for recommendation systems that predicts items based on historical data and reward prompts, avoiding online exploration errors.

Contribution

It proposes a new offline RL training approach for recommendation agents that directly infers actions from state-reward prompts, simplifying training with supervised learning.

Findings

PRL outperforms traditional RL methods on real-world datasets.

The approach simplifies offline training by using supervised learning.

Experimental results demonstrate superior recommendation accuracy.

Abstract

Modern recommender systems aim to improve user experience. As reinforcement learning (RL) naturally fits this objective -- maximizing an user's reward per session -- it has become an emerging topic in recommender systems. Developing RL-based recommendation methods, however, is not trivial due to the \emph{offline training challenge}. Specifically, the keystone of traditional RL is to train an agent with large amounts of online exploration making lots of `errors' in the process. In the recommendation setting, though, we cannot afford the price of making `errors' online. As a result, the agent needs to be trained through offline historical implicit feedback, collected under different recommendation policies; traditional RL algorithms may lead to sub-optimal policies under these offline training settings. Here we propose a new learning paradigm -- namely Prompt-Based Reinforcement Learning (PRL) -- for the offline training of RL-based recommendation agents. While traditional RL algorithms attempt to map state-action input pairs to their expected rewards (e.g., Q-values), PRL directly infers actions (i.e., recommended items) from state-reward inputs. In short, the agents are trained to predict a recommended item given the prior interactions and an observed reward value -- with simple supervised learning. At deployment time, this historical (training) data acts as a knowledge base, while the state-reward pairs are used as a prompt. The agents are thus used to answer the question: \emph{ Which item should be recommended given the prior interactions \& the prompted reward value}? We implement PRL with four notable recommendation models and conduct experiments on two real-world e-commerce datasets. Experimental results demonstrate the superior performance of our proposed methods.