Visually Dehallucinative Instruction Generation
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2024
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866914677635678208 |
|---|---|
| author | Cha, Sungguk Lee, Jusung Lee, Younghyun Yang, Cheoljong |
| author_facet | Cha, Sungguk Lee, Jusung Lee, Younghyun Yang, Cheoljong |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2402_08348 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | Visually Dehallucinative Instruction Generation Cha, Sungguk Lee, Jusung Lee, Younghyun Yang, Cheoljong Computer Vision and Pattern Recognition 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. |
| title | Visually Dehallucinative Instruction Generation |
| topic | Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2402.08348 |