Mariya Hendriksen

I am a Research Fellow at the University of Oxford, working with ʻŌiwi Parker Jones and Philip Torr. I am also a member of the ELLIS Society. I am passionate about advancing multimodal machine learning by creating models that are adaptable and trustworthy.

I earned my PhD in Artificial Intelligence at the University of Amsterdam, advised by Maarten de Rijke and Paul Groth. Before that, I obtained a MSc in AI from KU Leuven where I was advised by Marie-Francine Moens.

During my academic journey, I spent time at Microsoft Research, the Gemini team, Bloomberg AI, Amazon Science, LIIR at KU Leuven, and ETH Zurich as an intern. Alongside my research, I hold leadership roles in initiatives focused on diversity and inclusion in machine learning, including organizing the WiML Mentorship Program, serving as General Chair of WiML at ICML 2025, and mentoring through the ELLIS Inclusive AI initiative.

Email / Scholar / GitHub / Bsky / X

Updates

Publications

(Nov 2025) Presenting 'Adapting Vision-Language Models for Evaluating World Models' at NeurIPS LAW 2025 as an Oral.
(Sep 2025) Presenting 'Human-VLM Collaboration for Evaluating World Models' at CMU as a Spotlight.
(Sep 2025) Presenting 'Image-Seeking Intent Prediction for Cross-Device Product Search' at RecSys Gen AI for E-commerce 2025 as an Oral.
(Apr 2025) Paper 'Benchmark Granularity and Model Robustness for Image-Text Retrieval' accepted at SIGIR 2025.
(Jul 2024) Paper 'Demonstrating and Reducing Shortcuts in Vision-Language Representation Learning' accepted at TMLR.

Milestones

(Sep 2025) Started as a Research Fellow at the University of Oxford, working with Philip Torr and ʻŌiwi Parker Jones.
(Apr 2025) Serving as the General Chair for the Women in Machine Learning symposium at ICML 2025.
(Sep 2024) Started a research internship at Microsoft Research on the Game Intelligence team.
(Apr 2024) Started a research internship with the Gemini team.
(Jul 2023) Started a research internship at Bloomberg AI with the Question Answering team.

Research

	Adapting Vision-Language Models for Evaluating World Models. Mariya Hendriksen, Tabish Rashid, David Bignell, Raluca Georgescu, Abdelhak Lemkhenter, Katja Hoffman, Sam Devlin, Sarah Parisot Oral at NeurIPS LAW 2025; Under submission arXiv / GitHub We focus on the challenge of automated evaluation for world model rollouts and introduce a structured semantic protocol for this domain. We also propose UNIVERSE, a method for efficient adaptation of VLMs to this scenario.
	Demonstrating and Reducing Shortcuts in Vision-Language Representation Learning. Maurits Bleeker, Mariya Hendriksen, Andrew Yates, Maarten de Rijke TMLR, 2024 arXiv / bibtex / GitHub We propose a framework to examine the shortcut learning problem in the context of Vision-Language contrastive representation learning with multiple captions per image. We show how this problem can be partially mitigated using a form of text reconstruction and implicit feature modification.
	Multimodal Learned Sparse Retrieval with Probabilistic Expansion Control. Thong Nguyen, Mariya Hendriksen, Andrew Yates, Maarten de Rijke ECIR, 2024 arXiv / GitHub We propose a framework for multimodal learned sparse retrieval.
	Scene-Centric vs. Object-Centric Image-Text Cross-Modal Retrieval: A Reproducibility Study. Mariya Hendriksen, Svitlana Vakulenko, Ernst Kuiper, Maarten de Rijke ECIR, 2023 arXiv / GitHub
	Extending CLIP for Category-to-Image Retrieval in E-commerce . Mariya Hendriksen, Maurits Bleeker, Svitlana Vakulenko, Nanne van Noord, Maarten de Rijke ECIR, 2022 arXiv / GitHub
	Analyzing and Predicting Purchase Intent in E-commerce: Anonymous vs. Identified Customers. Mariya Hendriksen, Ernst Kuiper, Pim Nauts, Sebastian Schelter, Maarten de Rijke SIGIR eCom, 2020 arXiv / related work