Principal Research Scientist · London

Ismail Elezi, Ph.D.

Hi, I'm Ismail. I am a Principal Research Scientist at Huawei Noah's Ark in London, working with Jiankang Deng, where I am leading one of the multi-modality learning teams.

After finishing my bachelor studies at the University of Prishtina, Kosovo, I obtained both my master's and Ph.D. degree at Ca' Foscari University of Venice, Italy, advised by professors Marcello Pelillo and Thilo Stadelmann. Then, I worked as a postdoctoral researcher with professor Laura Leal-Taixé at Technical University of Munich and interned with Jose M. Alvarez at NVIDIA, Santa Clara.

Internship note. If you have a first author paper in a top ML/CV conference, and want to do an internship with me, please send me an email (find it in my CV). My interns/mentees usually author papers with me and go on to great companies or PhD programs.

View publicationsDownload CV

News

Recent papers, service, and personal research updates.

March 2026
One paper accepted to ICML 2026. Congratulations to Tatiana for MIDSteer.
March 2026
One paper accepted to CVPR 2026. Congratulations to Yura for HINT.
February 2026
Two papers accepted to ICLR 2026. Congratulations to Tatiana for CASteer and to the NLP team for DeepSynth. .
November 2025
One paper accepted to AAAI 2026. Congratulations to the team for ViCToR.
April 2025
Got chosen as Area Chair for NeurIPS 2025. First time I am doing this for a top-tier conference. #grownUp
February 2025
Got a paper accepted to CVPR 2025. Congratulations to Kostas.
January 2025
Finally got a paper accepted to ICLR. Congratulations to my former intern Prannay and see you in Singapore
September 2024
One paper accepted to NeurIPS 2024. Congratulations to Roy for VeLoRA. See you in Vancouver.
February 2024
Two papers accepted to CVPR 2024. Congratulations to Prady for G3DR and Roy for VKD! See you at Seattle.
November 2023
One paper got accepted to AAAI 2024. Congratulations to Chencheng!
April 2023
Started a new position as a Senior Research Scientist at Huawei Noah's Ark lab, working with Jiankang Deng. Very excited for the new work and for moving to London!
February 2023
One paper got accepted to CVPR 2023. Congratulations to Jenny!
September 2022
Two papers accepted to NeurIPS 2022. Congratulations to Vlad and Peter. Happy to be back to the US after covid.

Selected publications

Search by title, author, venue, or year. Publication cards keep the original links and thumbnails.

22 publications shown
Thumbnail for MIDSTEER: Optimal Affine Framework for Steering Generative Models
ICML2026

T. Gaintseva, A. Stepanov, Z. Liu, M. Benning, G. Slabaugh, J. Deng, I. Elezi: : MIDSTEER: Optimal Affine Framework for Steering Generative Models, International Conference on Machine Learning (ICML), 2026. paper

Thumbnail for Y. Choi, R. Miles, R. Potamias, I. Elezi , J. Deng, S. Zafeiriou, Gesture-Based Egocentric
Publication2026

Y. Choi, R. Miles, R. Potamias, I. Elezi, J. Deng, S. Zafeiriou, Gesture-Based Egocentric Video Question Answering , Conference on Computer Vision and Pattern Recognition (CVLR), 2026. paper

Thumbnail for CASteer: Cross-Attention Steering for Controllable Concept Erasure
ICLR2026

T. Gaintseva, A. Oncescu, C. Ma, Z. Liu, M. Benning, G. Slabaugh, J. Deng, I. Elezi: : CASteer: Cross-Attention Steering for Controllable Concept Erasure , International Conference on Learning Representations (ICLR), 2026.

Thumbnail for A Benchmark for Deep Information Synthesis
ICLR2026

D. Paul, D. Murphy, M. Gritta, R. Cardenas, V. Prokhorov, L. Bolliger, A. Toker, R. Miles, A. Oncescu, J. Sivakumar, P. Borchert, I. Elezi, M. Zhang, K. Lee, G. Zhang, G. Lampouras, and J. Wan: : A Benchmark for Deep Information Synthesis , International Conference on Learning Representations (ICLR), 2026. paper

Thumbnail for ViCToR: Improving Visual Comprehension via Token Reconstruction for Pretraining LMMs
AAAI2026

Y. Xie, K. Yang, P.Liang, X. An, Y.Zhao, Y. Wang, Z. Feng, R. Miles, I. Elezi, J. Deng: : ViCToR: Improving Visual Comprehension via Token Reconstruction for Pretraining LMMs , Conference on Artificial Intelligence (AAAI), 2026. paper

Thumbnail for “Principal Components” Enable A New Language of Images
ICCV2025

X. Wen, B. Zhao, I. Elezi, J. Deng, X. Qi : “Principal Components” Enable A New Language of Images , International Conference on Computer Vision (ICCV), 2025. paper

Thumbnail for Region-based Cluster Discrimination for Visual Representation Learning
ICCV2025

Y. Xie, K. Yang, X. An, K. Wu, Y. Zhao, W. Deng, Z. Ran, Y. Wang, Z. Feng, R. Miles, I. Elezi, J. Deng: Region-based Cluster Discrimination for Visual Representation Learning , International Conference on Computer Vision (ICCV), 2025. paper

Thumbnail for Fractal Calibration for long-tailed object detection
CVPR2025

K. Alexandridis, I. Elezi, J. Deng, A. Nguyen, S. Luo: Fractal Calibration for long-tailed object detection, Conference on Computer Vision and Pattern Recognition (CVPR), 2025. paper

Thumbnail for From Attention to Activation: Unraveling the Enigmas of Large Language Models
ICLR2025

P. Kaul, C. Ma, I. Elezi, J. Deng: From Attention to Activation: Unraveling the Enigmas of Large Language Models, International Conference on Learning Representations (ICLR), 2025. paper

Thumbnail for VeLoRA: Memory Efficient Training using Rank-1 Sub-Token Projections
NeurIPS2024

R. Miles, P.Reddy, I. Elezi, J. Deng: VeLoRA: Memory Efficient Training using Rank-1 Sub-Token Projections, Neural Information Processing Systems (NeurIPS), 2024. code paper

Thumbnail for G3DR: Generative 3D Reconstruction in ImageNet
CVPR2024

P. Reddy*, I. Elezi*, J. Deng: G3DR: Generative 3D Reconstruction in ImageNet, Conference on Computer Vision and Pattern Recognition (CVPR), 2024. page code paper

Thumbnail for VkD: Improving Knowledge Distillation using Orthogonal Projections
CVPR2024

R. Miles, I. Elezi, J. Deng: VkD: Improving Knowledge Distillation using Orthogonal Projections, Conference on Computer Vision and Pattern Recognition (CVPR), 2024. code paper

Thumbnail for Three Heads Are Better Than One: Complementary Experts for Long-Tailed Semi-supervised Learning
AAAI2024

C. Ma, I. Elezi, J. Deng, W. Dong, C. Xu: Three Heads Are Better Than One: Complementary Experts for Long-Tailed Semi-supervised Learning, Conference on Artificial Intelligence (AAAI), 2024. code paper

Thumbnail for Simple Cues Lead to a Strong Multi-Object Tracker
CVPR2023

J. Seidenschwarz, G. Braso, V. Serrano, I. Elezi, L. Leal-Taixé: Simple Cues Lead to a Strong Multi-Object Tracker, Conference on Computer Vision and Pattern Recognition (CVPR), 2023. code paper video

Thumbnail for The group loss++: A deeper look into group loss for deep metric learning
PAMI2023

I. Elezi*, J. Seidenschwarz*, L. Wagner*, S. Vascon, A. Torcinovich, M. Pelillo, L. Leal-Taixé: The group loss++: A deeper look into group loss for deep metric learning, Transactions on Pattern Analysis and Machine Intelligence (PAMI), 2023. paper

Thumbnail for Learning to Discover and Detect Objects
NeurIPS2022

V. Fomenko, I. Elezi, D. Ramanan, L. Leal-Taixé, A. Ošep: Learning to Discover and Detect Objects, Neural Information Processing Systems (NeurIPS), 2022. page poster video code paper

Thumbnail for The Unreasonable Effectiveness of Fully-Connected Layers for Low-Data Regimes
NeurIPS2022

P. Kocsis, P. Sukenik, G. Braso, M. Niessner, L. Leal-Taixé, I. Elezi: The Unreasonable Effectiveness of Fully-Connected Layers for Low-Data Regimes, Neural Information Processing Systems (NeurIPS), 2022. page code paper

Thumbnail for Not All Labels Are Equal: Rationalizing The Labeling Costs for Training Object Detection
CVPR2022

I. Elezi, Z. Yu, A. Anandkumar, L. Leal-Taixé, J. Alvarez: Not All Labels Are Equal: Rationalizing The Labeling Costs for Training Object Detection, Conference on Computer Vision and Pattern Recognition (CVPR), 2022. code paper

Thumbnail for Active Learning for Deep Object Detection via Probabilistic Modeling
ICCV2021

J. Choi, I. Elezi, H. Lee, C. Farabet, J. Alvarez: Active Learning for Deep Object Detection via Probabilistic Modeling, International Conference on Computer Vision (ICCV), 2021. code paper video

Thumbnail for Learning Intra-Batch Connections for Deep Metric Learning
ICML2021

J. Seidenschwarz, I. Elezi, L. Leal-Taixé: Learning Intra-Batch Connections for Deep Metric Learning, International Conference on Machine Learning (ICML), 2021. code paper video

Thumbnail for The Group Loss for Deep Metric Learning
ECCV2020

I. Elezi, S. Vascon, A. Torcinovich, M. Pelillo, L. Leal-Taixé: The Group Loss for Deep Metric Learning, European Conference on Computer Vision (ECCV), 2020. page code paper video

Thumbnail for CIAGAN: Conditional Identity Anonymization Generative Adversarial Networks
CVPR2020

M. Maximov*, I. Elezi*, L. Leal-Taixé: CIAGAN: Conditional Identity Anonymization Generative Adversarial Networks, Conference on Computer Vision and Pattern Recognition (CVPR), 2020. code paper video

Students and interns

People supervised and mentored across internships, theses, and research projects.

Service

Conference service, reviewing, awards, and community contributions.

Teaching

Courses, lectures, and public teaching material.

Winter 2021
Lecturer for IIN2375: Computer Vision III: Detection, Segmentation and Tracking (CV3DST) (TU Munich)
  • Winter 2021: Lecturer for IN2389: Advanced Deep Learning for Computer vision (ADL4CV) (TU Munich)
  • 2019-2023: Lecturer for Introduction to Deep Learning with PyTorch (Datacamp). Around 30K students took the course before it got retired in 2023.
  • I have a few lectures online in Transformers, Semi-Supervised Learning, and Active Learning. They are a couple of years old so relatively outdated.