ELLIS-ELLIOT Reading Group: Human-Centric Machine Learning Session
This reading group aims to gather researchers and students interested in both getting a wide vision of the topic and also deeply diving into it. Reading papers about different topics inside HCML, and also discussing new problem set-ups, different approaches, and sources of bias will lead us to a broad understanding of how algorithmic and human decisions influence each other. It's organised by students in the ELLIS PhD Program and the ELLIOT Young Researcher Group.
Learn more about the Reading Group or join the Google Group to stay informed about the session details.
Session Topic
Title: Vision-Language Models Do Not Understand Negation
Abstract: Many practical vision-language applications require models that understand negation, e.g., when using natural language to retrieve images which contain certain objects but not others. Despite advancements in vision-language models (VLMs) through large-scale training, their ability to comprehend negation remains underexplored. This study addresses the question: how well do current VLMs understand negation? We introduce NegBench, a new benchmark designed to evaluate negation understanding across 18 task variations and 79k examples spanning image, video, and medical datasets. The benchmark consists of two core tasks designed to evaluate negation understanding in diverse multimodal settings: Retrieval with Negation and Multiple Choice Questions with Negated Captions. Our evaluation reveals that modern VLMs struggle significantly with negation, often performing at chance level. To address these shortcomings, we explore a data-centric approach wherein we finetune CLIP models on large-scale synthetic datasets containing millions of negated captions. We show that this approach can result in a 10% increase in recall on negated queries and a 28% boost in accuracy on multiple-choice questions with negated captions.