Dr. Kovashka received her BA degrees in Computer Science and Media Studies from Pomona College, in 2008, and her PhD in Computer Science from The University of Texas in Austin, in 2014. She joined Pitt's Computer Science department in January 2015.
Her primary research area is computer vision, with overlap in machine learning and natural language processing. One thrust in her work is examination of distribution shifts affecting models, arising due to geographic, demographic, and stylistic factors. Another thrust focuses on computational models of visual persuasion, including automated understanding and generation of visual advertisements. A third focus is on advancing computer vision models for understanding and supporting human learning.
- Role Bias in Text-to-Image Diffusion Models: Diagnosing and Mitigating Compositional Failures through Intermediate Decomposition. Sina Malakouti and Adriana Kovashka. In Neural Information Processing Systems (NeurIPS), December 2025.
- Probing Logical Reasoning of MLLMs in Scientific Diagrams. Yufei (Fiona) Wang and Adriana Kovashka. In Empirical Methods in Natural Language Processing (EMNLP) Main Conference (Short), November 2025.
- CAP: Evaluation of Persuasive and Creative Image Generation. Aysan Aghazadeh and Adriana Kovashka. In International Conference on Computer Vision (ICCV), October 2025.
- Towards Understanding Ambiguity Resolution in Multimodal Inference of Meaning. Yufei (Fiona) Wang, Adriana Kovashka, Loretta Fernandez, Marc N. Coutanche, Seth Wiener. In International Conference on Development and Learning (ICDL), September 2025.
- Incorporating Geo-Diverse Knowledge into Prompting for Increased Geographical Robustness in Object Recognition. Kyle Buettner, Sina Malakouti, Xiang Lorraine Li, Adriana Kovashka. In IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), June 2024.
Computer Vision
Natural Language Processing
Domain Generalization
Visual Persuasion
New NSF grant on building AI models to support students interpret scientific diagrams, in collaboration with Indiana University Bloomington
NSF grant to study geographic diversity in visual/context appearance, language supervision, and object vocabularies, for object detection
NSF CAREER for developing methods to handle noise and ambiguity in weak language and multimodal supervision for object detection
