KnowRL: Exploring Knowledgeable Reinforcement Learning for Factuality

TL;DR

KnowRL introduces a factuality reward mechanism into reinforcement learning to reduce hallucinations in large language models during reasoning, improving factual accuracy without sacrificing reasoning ability.

Contribution

It presents a novel knowledge-enhanced RL approach that guides models to perform fact-based slow thinking, addressing hallucination issues in LLMs.

Findings

KnowRL significantly reduces hallucinations across multiple datasets.

The method maintains the original reasoning capabilities of models.

Factuality rewards improve the recognition of knowledge boundaries during reasoning.

Abstract

Large Language Models (LLMs), particularly slow-thinking models, often exhibit severe hallucination, outputting incorrect content due to an inability to accurately recognize knowledge boundaries during reasoning. While Reinforcement Learning (RL) can enhance complex reasoning abilities, its outcome-oriented reward mechanism often lacks factual supervision over the thinking process, further exacerbating the hallucination problem. To address the high hallucination in slow-thinking models, we propose Knowledge-enhanced RL, KnowRL. KnowRL guides models to perform fact-based slow thinking by integrating a factuality reward, based on knowledge verification, into the RL training process, helping them recognize their knowledge boundaries. KnowRL guides models to perform fact-based slow thinking by integrating a factuality reward, based on knowledge verification, into the RL training process, helping them recognize their knowledge boundaries. This targeted factual input during RL training enables the model to learn and internalize fact-based reasoning strategies. By directly rewarding adherence to facts within the reasoning steps, KnowRL fosters a more reliable thinking process. Experimental results on three hallucination evaluation datasets and two reasoning evaluation datasets demonstrate that KnowRL effectively mitigates hallucinations in slow-thinking models while maintaining their original strong reasoning capabilities. Our code is available at https://github.com/zjunlp/KnowRL.