Large language models readily adapt to novel settings, even without task-specific training data. Can their zero-shot capacity be extended to multimodal inputs? In this work, we propose ESPER which extends language-only zero-shot models to unseen multimodal tasks, like image and audio captioning. Our key novelty is to use reinforcement learning to align multimodal inputs to language model generations without direct supervision: for example, in the image case our reward optimization relies only on cosine similarity derived from CLIP, and thus requires no additional explicitly paired (image, caption) data. Because the parameters of the language model are left unchanged, the model maintains its capacity for zero-shot generalization. Experiments demonstrate that ESPER outperforms baselines and prior work on a variety of zero-shot tasks; these include a new benchmark we collect+release, ESP dataset, which tasks models with generating several diversely-styled captions for each image.
@article{arxiv.2205.12630,
title = {Multimodal Knowledge Alignment with Reinforcement Learning},
author = {Youngjae Yu and Jiwan Chung and Heeseung Yun and Jack Hessel and JaeSung Park and Ximing Lu and Prithviraj Ammanabrolu and Rowan Zellers and Ronan Le Bras and Gunhee Kim and Yejin Choi},
journal= {arXiv preprint arXiv:2205.12630},
year = {2022}
}