Visually Dehallucinative Instruction Generation: Know What You Don't Know
This addresses the challenge of generating accurate 'I Don't Know' responses for unanswerable image-question pairs in vision-language models, representing an incremental improvement in a specific domain.
The paper tackles the problem of visual hallucination in generative language models by introducing the 'I Know' hallucination concept and proposing a visually dehallucinative instruction generation method, which effectively reduces these hallucinations across different frameworks and datasets.
"When did the emperor Napoleon invented iPhone?" Such hallucination-inducing question is well known challenge in generative language modeling. In this study, we present an innovative concept of visual hallucination, referred to as "I Know (IK)" hallucination, to address scenarios where "I Don't Know" is the desired response. To effectively tackle this issue, we propose the VQAv2-IDK benchmark, the subset of VQAv2 comprising unanswerable image-question pairs as determined by human annotators. Stepping further, we present the visually dehallucinative instruction generation method for IK hallucination and introduce the IDK-Instructions visual instruction database. Our experiments show that current methods struggle with IK hallucination. Yet, our approach effectively reduces these hallucinations, proving its versatility across different frameworks and datasets.