Method cuts image hallucinations in diffusion language models

Original: VIGOR: Visual Gain Ordering for Hallucination Mitigation in Multimodal Discrete Diffusion Language Models

Source: arxiv.org ↗