Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Singh, Ashish, Singh, Ashutosh, Agarwal, Prateek, Huang, Zixuan, Singh, Arpita, Yu, Tong, Kim, Sungchul, Bursztyn, Victor, Ahmed, Nesreen K., Mathur, Puneet, Learned-Miller, Erik, Dernoncourt, Franck, Rossi, Ryan A.
Format: Preprint
Veröffentlicht: 2023
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2307.10867
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913896805171200
author Singh, Ashish
Singh, Ashutosh
Agarwal, Prateek
Huang, Zixuan
Singh, Arpita
Yu, Tong
Kim, Sungchul
Bursztyn, Victor
Ahmed, Nesreen K.
Mathur, Puneet
Learned-Miller, Erik
Dernoncourt, Franck
Rossi, Ryan A.
author_facet Singh, Ashish
Singh, Ashutosh
Agarwal, Prateek
Huang, Zixuan
Singh, Arpita
Yu, Tong
Kim, Sungchul
Bursztyn, Victor
Ahmed, Nesreen K.
Mathur, Puneet
Learned-Miller, Erik
Dernoncourt, Franck
Rossi, Ryan A.
contents Captions are crucial for understanding scientific visualizations and documents. Existing captioning methods for scientific figures rely on figure-caption pairs extracted from documents for training, many of which fall short with respect to metrics like helpfulness, explainability, and visual-descriptiveness [15] leading to generated captions being misaligned with reader preferences. To enable the generation of high-quality figure captions, we introduce FigCaps-HF a new framework for figure-caption generation that can incorporate domain expert feedback in generating captions optimized for reader preferences. Our framework comprises of 1) an automatic method for evaluating quality of figure-caption pairs, 2) a novel reinforcement learning with human feedback (RLHF) method to optimize a generative figure-to-caption model for reader preferences. We demonstrate the effectiveness of our simple learning framework by improving performance over standard fine-tuning across different types of models. In particular, when using BLIP as the base model, our RLHF framework achieves a mean gain of 35.7%, 16.9%, and 9% in ROUGE, BLEU, and Meteor, respectively. Finally, we release a large-scale benchmark dataset with human feedback on figure-caption pairs to enable further evaluation and development of RLHF techniques for this problem.
format Preprint
id arxiv_https___arxiv_org_abs_2307_10867
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle FigCaps-HF: A Figure-to-Caption Generative Framework and Benchmark with Human Feedback
Singh, Ashish
Singh, Ashutosh
Agarwal, Prateek
Huang, Zixuan
Singh, Arpita
Yu, Tong
Kim, Sungchul
Bursztyn, Victor
Ahmed, Nesreen K.
Mathur, Puneet
Learned-Miller, Erik
Dernoncourt, Franck
Rossi, Ryan A.
Computation and Language
Computer Vision and Pattern Recognition
Machine Learning
Captions are crucial for understanding scientific visualizations and documents. Existing captioning methods for scientific figures rely on figure-caption pairs extracted from documents for training, many of which fall short with respect to metrics like helpfulness, explainability, and visual-descriptiveness [15] leading to generated captions being misaligned with reader preferences. To enable the generation of high-quality figure captions, we introduce FigCaps-HF a new framework for figure-caption generation that can incorporate domain expert feedback in generating captions optimized for reader preferences. Our framework comprises of 1) an automatic method for evaluating quality of figure-caption pairs, 2) a novel reinforcement learning with human feedback (RLHF) method to optimize a generative figure-to-caption model for reader preferences. We demonstrate the effectiveness of our simple learning framework by improving performance over standard fine-tuning across different types of models. In particular, when using BLIP as the base model, our RLHF framework achieves a mean gain of 35.7%, 16.9%, and 9% in ROUGE, BLEU, and Meteor, respectively. Finally, we release a large-scale benchmark dataset with human feedback on figure-caption pairs to enable further evaluation and development of RLHF techniques for this problem.
title FigCaps-HF: A Figure-to-Caption Generative Framework and Benchmark with Human Feedback
topic Computation and Language
Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2307.10867