Revisiting the Role of Language Priors in Vision-Language Models

Zhiqiu Lin, Xinyue Chen, Deepak Pathak, Pengchuan Zhang, Deva Ramanan

Keywords: Vision-Language Models, Visio-Linguistic Compositionality, Cross-Modal Retrieval, Computer Vision, Natural Language Processing

TL;DR: We study generative VLMs for image-text retrieval tasks, achieving SOTA performance via debiasing of generative scores without any fine-tuning.

Abstract: Vision-language models (VLMs) are impactful in part because they can be applied to a variety of visual understanding tasks in a zero-shot fashion, without any fine-tuning. We study currently popular *generative VLMs* that are trained for next-word generation given the image. We explore their zero-shot performance on the illustrative task of image-text retrieval across 8 popular vision-language benchmarks. Our first observation is that they can be repurposed for discriminative tasks (such as image-text retrieval) by simply computing the match score of generating a particular text string given an image. We call this probabilistic score the *Visual Generative Pre-Training Score* (VisualGPTScore). While the VisualGPTScore produces near-perfect accuracy on some retrieval benchmarks, it produces poor accuracy on others. We analyze this behavior through a probabilistic lens, pointing out that some benchmarks inadvertently capture unnatural language distributions by creating adversarial but unlikely text captions. In fact, we demonstrate that even a ``blind'' language model that ignores any image evidence can sometimes outperform all prior art, reminiscent of similar challenges faced by the visual-question answering (VQA) community many years ago. We derive a probabilistic post-processing scheme that controls for the amount of linguistic bias in generative VLMs at test time without having to retrain or fine-tune the model. We show that the VisualGPTScore, when appropriately debiased, is a strong zero-shot baseline for vision-language understanding, oftentimes producing state-of-the-art accuracy.

Primary Area: representation learning for computer vision, audio, language, and other modalities

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.

Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2024/AuthorGuide.

Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors' identity.

No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.

Submission Number: 721