Vision-language (VL) models, pretrained on colossal image-text datasets, have attained broad VL competence that is difficult to evaluate. A common belief is that a small number of VL skills underlie the variety of VL tests. In this paper, we perform a large-scale transfer learning experiment aimed at discovering latent VL skills from data. We reveal interesting characteristics that have important implications for test suite design. First, generation tasks suffer from a length bias, suggesting benchmarks should balance tasks with varying output lengths. Second, we demonstrate that factor analysis successfully identifies reasonable yet surprising VL skill factors, suggesting benchmarks could leverage similar analyses for task selection. Finally, we present a new dataset, OLIVE (https://github.com/jq-zh/olive-dataset), which simulates user instructions in the wild and presents challenges dissimilar to all datasets we tested. Our findings contribute to the design of balanced and broad-coverage vision-language evaluation methods.
Ask Me Again Differently: GRAS for Measuring Bias in Vision Language Models on Gender, Race, Age, and Skin Tone
Guru Gobind Singh Indraprastha University（古鲁·戈宾德·辛格印度教普拉斯塔大学）；AI Institute, University of South Carolina（南卡罗来纳大学人工智能研究所）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Indian Institute of Technology Patna, India（印度理工学院帕纳布分校）
VisBias: Measuring Explicit and Implicit Social Biases in Vision Language Models
University of Southern California（南加州大学）；University of Bristol（布里斯托大学）；Centre National de la Recherche Scientifique（法国国家科学研究中心）；University of California, Los Angeles（加州大学洛杉矶分校）
Machine learning models are known to perpetuate and even amplify the biases present in the data. However, these data biases frequently do not become apparent until after the models are deployed. Our work tackles this issue and enables the preemptive analysis of large-scale datasets. REVISE (REvealing VIsual biaSEs) is a tool that assists in the investigation of a visual dataset, surfacing potential biases along three dimensions: (1) object-based, (2) person-based, and (3) geography-based. Object-based biases relate to the size, context, or diversity of the depicted objects. Person-based metric…
University of Edinburgh（爱丁堡大学）；IMT School for Advanced Studies Lucca（卢卡高等研究院）；Canon Medical Research Europe Ltd.（佳能医疗欧洲研究有限公司）；The Alan Turing Institute（艾伦·图灵研究所）
A recent spate of state-of-the-art semi- and un-supervised solutions disentangle and encode image "content" into a spatial tensor and image appearance or "style" into a vector, to achieve good performance in spatially equivariant tasks (e.g. image-to-image translation). To achieve this, they employ different model design, learning objective, and data biases. While considerable effort has been made to measure disentanglement in vector representations, and assess its impact on task performance, such analysis for (spatial) content - style disentanglement is lacking. In this paper, we conduct an e…
Google（谷歌）；Ulm University（乌尔姆大学）；Ethical AI LLC（Ethical AI有限责任公司）
The measurement of bias in machine learning often focuses on model performance across identity subgroups (such as man and woman) with respect to groundtruth labels. However, these methods do not directly measure the associations that a model may have learned, for example between labels and identity subgroups. Further, measuring a model's bias requires a fully annotated evaluation dataset which may not be easily available in practice. We present an elegant mathematical solution that tackles both issues simultaneously, using image classification as a working example. By treating a classification…
Latent Representation Matters: Human-like Sketches in One-shot Drawing Tasks
Artificial and Natural Intelligence Toulouse Institute（图卢兹人工与自然智能研究所）；Université de Toulouse（图卢兹大学）；Centre de Recherche Cerveau & Cognition CNRS（法国国家科学研究中心大脑与认知研究中心）；Carney Institute for Brain Science, Brown University（布朗大学卡尼脑科学研究所）
Deep learning (DL) models are widely used to provide a more convenient and smarter life. However, biased algorithms will negatively influence us. For instance, groups targeted by biased algorithms will feel unfairly treated and even fearful of negative consequences of these biases. This work targets biased generative models' behaviors, identifying the cause of the biases and eliminating them. We can (as expected) conclude that biased data causes biased predictions of face frontalization models. Varying the proportions of male and female faces in the training data can have a substantial effect …
From Hope to Safety: Unlearning Biases of Deep Models via Gradient Penalization in Latent Space
Fraunhofer Heinrich Hertz Institut（弗劳恩霍夫海因里希·赫兹研究所）；Technische Universität Berlin（柏林工业大学）；Berlin Institute for the Foundations of Learning and Data (BIFOLD)（柏林学习与数据基础研究所（BIFOLD））
GeoDiv: Framework For Measuring Geographical Diversity In Text-To-Image Models
Vision and AI Lab, Indian Institute of Science, Bangalore, India（印度科学院视觉与人工智能实验室）；Chair for Machine Learning, University of Mannheim（曼海姆大学机器学习主任）；Max Planck Institute for Informatics, Saarland Informatics Campus, Saarbrücken, Germany（马克斯·普朗克信息研究所，萨尔兰州信息学院，萨尔布吕肯，德国）
Arizona State University（亚利桑那州立大学）；Rensselaer Polytechnic Institute（伦斯勒理工学院）
In this paper, we show that popular Generative Adversarial Networks (GANs) exacerbate biases along the axes of gender and skin tone when given a skewed distribution of face-shots. While practitioners celebrate synthetic data generation using GANs as an economical way to augment data for training data-hungry machine learning models, it is unclear whether they recognize the perils of such techniques when applied to real world datasets biased along latent dimensions. Specifically, we show that (1) traditional GANs further skew the distribution of a dataset consisting of engineering faculty headsh…
