GFlowGR: Fine-tuning Generative Recommendation Frameworks with Generative Flow Networks
Researchers have introduced GFlowGR, a novel fine-tuning framework for generative recommendation systems that utilizes Generative Flow Networks (GFlowNets). This approach aims to address the exposure bias problem inherent in current methods like supervised fine-tuning (SFT) and direct preference optimization (DPO). GFlowGR integrates collaborative knowledge from traditional recommender systems to create an adaptive sampler and reward model, thereby mitigating exposure bias and improving recommendation quality. AI
IMPACT This research introduces a novel method to improve generative recommendation systems by addressing exposure bias, potentially leading to more accurate and diverse recommendations.