Visually Dehallucinative Instruction Generation
This addresses the issue of unintended content in synthetic visual instructions for researchers and practitioners in AI, though it appears incremental as it builds on existing methods for visual instruction tuning.
The paper tackles the problem of hallucination in generative language models for visual question-answering by introducing CAP2QA, a method that constrains generation to image contents, resulting in significantly reduced visual hallucination and improved visual recognition ability and expressiveness.
In recent years, synthetic visual instructions by generative language model have demonstrated plausible text generation performance on the visual question-answering tasks. However, challenges persist in the hallucination of generative language models, i.e., the generated image-text data contains unintended contents. This paper presents a novel and scalable method for generating visually dehallucinative instructions, dubbed CAP2QA, that constrains the scope to only image contents. Our key contributions lie in introducing image-aligned instructive QA dataset CAP2QA-COCO and its scalable recipe. In our experiments, we compare synthetic visual instruction datasets that share the same source data by visual instruction tuning and conduct general visual recognition tasks. It shows that our proposed method significantly reduces visual hallucination while consistently improving visual recognition ability and expressiveness.