Abstract: Multimodal Large Language Models (MLLMs) frequently suffer from hallucination issues, generating information about objects that are not present in input images during vision-language tasks.
Some results have been hidden because they may be inaccessible to you
Show inaccessible results