Welcome! I'm Adheesh Juvekar
PhD Student at University of Illinois at Urbana-Champaign
I'm a Computer Science PhD student at PLAN Lab, UIUC, advised by Dr. Ismini Lourentzou. My research interests are in Multimodal Learning and Generative AI, with a focus on building models that can understand and create visual content in ways that are grounded, controllable, and interpretable.
I’m open to collaborations and always interested in discussing new research ideas, feel free to reach out via email. If you’re a self-motivated undergraduate (especially at UIUC) interested in working on Generative AI / multimodal projects feel free to reach out via email.
Research Areas
Multimodal Learning & Generative AI
My research focuses on how models connect language with visual evidence for multimodal understanding, generation, and editing. My current work investigates how models interpret visual regions and use regional information to guide reasoning and control visual content. Through this work, I aim to develop models that understand scene structure and support precise, temporally consistent visual generation and editing. I'm also interested in reasoning at the intersection of generation and perception, including spatial and temporal reasoning, 3D-consistent generation, and unified models for understanding, generation, and editing, as a step towards reliable and robust world models.
Publications & Preprints
JOLT: Uncertainty-Selective Self-Ensembling for Geometric Consistency in Video Generation
From Tokens to Trajectories: Parameter-Efficient Continuous Adaptation of Unified Multimodal Models with FlowShift
Fluxlearn’s to Write
UniSegFlow: Segmentation as a Unified Thinking Paradigm for Visual Understanding, Generation, and Editing via Discrete Flow Matching
PRIMA: Multi-Image Vision-Language Models for Reasoning Segmentation
* Equal contribution