Personalized Scientific Figure Caption Generation: An Empirical Study on Author-Specific Writing Style Transfer
This work addresses the problem of automating scientific figure captions with author-specific styles for researchers and publishers, but it is incremental as part of a challenge.
The study tackled personalized figure caption generation by using author profile data and metadata to improve personalization in multimodal large language models, revealing a trade-off between matching author style and maintaining caption quality.
We study personalized figure caption generation using author profile data from scientific papers. Our experiments demonstrate that rich author profile data, combined with relevant metadata, can significantly improve the personalization performance of multimodal large language models. However, we also reveal a fundamental trade-off between matching author style and maintaining caption quality. Our findings offer valuable insights and future directions for developing practical caption automation systems that balance both objectives. This work was conducted as part of the 3rd SciCap challenge.