KnowRL: Exploring Knowledgeable Reinforcement Learning for Factuality
This addresses the issue of factual inaccuracies in AI-generated content for users relying on reliable language models, representing an incremental improvement by enhancing existing RL methods with targeted factual supervision.
The paper tackles the problem of severe hallucination in slow-thinking large language models by proposing Knowledge-enhanced RL (KnowRL), which integrates a factuality reward based on knowledge verification into RL training to guide fact-based reasoning, and experimental results on three hallucination and two reasoning datasets show that KnowRL effectively mitigates hallucinations while maintaining strong reasoning capabilities.
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.