Accumulative Decoding: An Innovative Method to Reduce Hallucinations in Vision-Language Models Without Training

Introduction / Main Post: Accumulative Decoding: An Innovative Method to Reduce Hallucinations in Vision-Language Models Without Training

Accumulative-Decoding is an innovative open-source project focused on addressing the hallucination problem in large Vision-Language Models (VLMs). The project proposes an accumulative decoding method that reduces model hallucinations without additional training, improving the accuracy and reliability of model outputs by enhancing the decoding strategy.