SK-VQA: Synthetic Knowledge Generation at Scale for Training Context-Augmented Multimodal LLMs

Xin Su, Man Luo, Kris W Pan, Tien Pei Chou, Vasudev Lal, Phillip Howard
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:57008-57029, 2025.

Abstract

Multimodal retrieval-augmented generation (RAG) plays a crucial role in domains such as knowledge-based visual question answering (KB-VQA), where models should effectively integrate additional knowledge to generate a response. However, existing vision and language models (VLMs) are not inherently designed for context-augmented generation, limiting their effectiveness in such tasks. While synthetic data generation has recently gained attention for training large VLMs, its application for context-augmented generation remains underexplored. To address this gap, we introduce SKVQA, a large-scale synthetic multimodal dataset containing over 2 million visual question-answer pairs, each associated with external knowledge sources to determine the final answer. Compared to previous datasets, SKVQA exhibits 11$\times$ more unique questions, greater domain diversity, and a broader spectrum of image sources. Through human evaluations, we confirm the high quality of the generated question-answer pairs and their contextual relevance. Extensive experiments show that SKVQA serves both as a challenging benchmark for knowledge-based VQA and as an effective training resource for adapting generative multimodal models to context-augmented generation. Our results further indicate that models trained on SKVQA demonstrate enhanced generalization in both context-aware VQA and multimodal RAG settings.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-su25a, title = {{SK}-{VQA}: Synthetic Knowledge Generation at Scale for Training Context-Augmented Multimodal {LLM}s}, author = {Su, Xin and Luo, Man and Pan, Kris W and Chou, Tien Pei and Lal, Vasudev and Howard, Phillip}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {57008--57029}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/su25a/su25a.pdf}, url = {https://proceedings.mlr.press/v267/su25a.html}, abstract = {Multimodal retrieval-augmented generation (RAG) plays a crucial role in domains such as knowledge-based visual question answering (KB-VQA), where models should effectively integrate additional knowledge to generate a response. However, existing vision and language models (VLMs) are not inherently designed for context-augmented generation, limiting their effectiveness in such tasks. While synthetic data generation has recently gained attention for training large VLMs, its application for context-augmented generation remains underexplored. To address this gap, we introduce SKVQA, a large-scale synthetic multimodal dataset containing over 2 million visual question-answer pairs, each associated with external knowledge sources to determine the final answer. Compared to previous datasets, SKVQA exhibits 11$\times$ more unique questions, greater domain diversity, and a broader spectrum of image sources. Through human evaluations, we confirm the high quality of the generated question-answer pairs and their contextual relevance. Extensive experiments show that SKVQA serves both as a challenging benchmark for knowledge-based VQA and as an effective training resource for adapting generative multimodal models to context-augmented generation. Our results further indicate that models trained on SKVQA demonstrate enhanced generalization in both context-aware VQA and multimodal RAG settings.} }
Endnote
%0 Conference Paper %T SK-VQA: Synthetic Knowledge Generation at Scale for Training Context-Augmented Multimodal LLMs %A Xin Su %A Man Luo %A Kris W Pan %A Tien Pei Chou %A Vasudev Lal %A Phillip Howard %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-su25a %I PMLR %P 57008--57029 %U https://proceedings.mlr.press/v267/su25a.html %V 267 %X Multimodal retrieval-augmented generation (RAG) plays a crucial role in domains such as knowledge-based visual question answering (KB-VQA), where models should effectively integrate additional knowledge to generate a response. However, existing vision and language models (VLMs) are not inherently designed for context-augmented generation, limiting their effectiveness in such tasks. While synthetic data generation has recently gained attention for training large VLMs, its application for context-augmented generation remains underexplored. To address this gap, we introduce SKVQA, a large-scale synthetic multimodal dataset containing over 2 million visual question-answer pairs, each associated with external knowledge sources to determine the final answer. Compared to previous datasets, SKVQA exhibits 11$\times$ more unique questions, greater domain diversity, and a broader spectrum of image sources. Through human evaluations, we confirm the high quality of the generated question-answer pairs and their contextual relevance. Extensive experiments show that SKVQA serves both as a challenging benchmark for knowledge-based VQA and as an effective training resource for adapting generative multimodal models to context-augmented generation. Our results further indicate that models trained on SKVQA demonstrate enhanced generalization in both context-aware VQA and multimodal RAG settings.
APA
Su, X., Luo, M., Pan, K.W., Chou, T.P., Lal, V. & Howard, P.. (2025). SK-VQA: Synthetic Knowledge Generation at Scale for Training Context-Augmented Multimodal LLMs. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:57008-57029 Available from https://proceedings.mlr.press/v267/su25a.html.

Related Material