Faegheh Sardari

ASG Office,

Microsoft Campus,

Thames Valley Park,

Reading, England, UK

Email: fsardari(@)microsoft(dot)com

Hello. Thanks for visiting my webpage! My name is Faegheh (Fay), and I’m a Senior Scientist at Microsoft, where I lead the multimodal post-training and alignment for Aion Instruct, a language model developed from scratch by our team at Microsoft ASG.

Prior to that, I was a Postdoctoral Research Fellow at CVSSP, University of Surrey, working with Prof. Adrian Hilton on AI4me project. Our goal was to leverage AI for personalised media production.

I completed my PhD at VI-Lab (now MaVi), University of Bristol, where I was awarded the university’s most prestigious and competitive PhD scholarship. During my PhD, I was supervised by Prof. Majid Mirmehdi and had the opportunity to collaborate with Prof. Björn Ommer from Ludwig Maximilian University of Munich.

News

Sep 25, 2026	`Latent-Lens’ is accepted at NeurIPS-2026, main-track.
Apr 01, 2026	`DocSLM’ is accepted at CVPR-2026, Findings.
Jan 01, 2026	Our paper `An Effective-Efficient Approach for Dense Multi-Label Action Detection’ is accepted at IJCV.
Oct 01, 2025	I joined Microsoft as a Senior Scientist.
Feb 27, 2025	We secured a £100K grant (50% funded by EPSRC and 50% by Tesco) to advance video understanding for retail.

Selected Publications

Effective-Efficient

An Effective-Efficient Approach for Dense Multi-Label Action Detection

Faegheh Sardari, Armin Mustafa, Philip JB Jackson, and Adrian Hilton

IJCV, 2025

arXiv Bib

@article{sardari2024effective,
  title = {An Effective-Efficient Approach for Dense Multi-Label Action Detection},
  author = {Sardari, Faegheh and Mustafa, Armin and Jackson, Philip JB and Hilton, Adrian},
  journal = {IJCV},
  year = {2025}
}

RefDense

Reframing Dense Action Detection (RefDense): A Paradigm Shift in Problem Solving & a Novel Optimization Strategy

Faegheh Sardari, Armin Mustafa, Philip JB Jackson, and Adrian Hilton

arXiv, 2025

arXiv Bib HTML

@article{sardari2025deconstruct,
  title = {Reframing Dense Action Detection (RefDense): A Paradigm Shift in Problem Solving & a Novel Optimization Strategy},
  author = {Sardari, Faegheh and Mustafa, Armin and Jackson, Philip JB and Hilton, Adrian},
  journal = {arXiv},
  year = {2025}
}

Narrativebridge

Narrativebridge: Enhancing Video Captioning with Causal-Temporal Narrative

Asmar Nadeem, Faegheh Sardari, Robert Dawes, Syed Sameed Husain, and 2 more authors

In ICLR, 2025

arXiv Bib HTML

@inproceedings{nadeem2024narrativebridge,
  title = {Narrativebridge: Enhancing Video Captioning with Causal-Temporal Narrative},
  author = {Nadeem, Asmar and Sardari, Faegheh and Dawes, Robert and Husain, Syed Sameed and Hilton, Adrian and Mustafa, Armin},
  booktitle = {ICLR},
  year = {2025}
}

CoLeaF

CoLeaF: A Contrastive-Collaborative Learning Framework for Weakly Supervised Audio-Visual Video Parsing

Faegheh Sardari, Armin Mustafa, Philip JB Jackson, and Adrian Hilton

In ECCV, 2024

arXiv Bib HTML

@inproceedings{sardari2024coleaf,
  title = {CoLeaF: A Contrastive-Collaborative Learning Framework for Weakly Supervised Audio-Visual Video Parsing},
  author = {Sardari, Faegheh and Mustafa, Armin and Jackson, Philip JB and Hilton, Adrian},
  booktitle = {ECCV},
  pages = {1--17},
  year = {2024},
  organization = {Springer}
}

ForecasterFlexOBM

ForecasterFlexOBM: A Multi-View Audio-Visual Dataset for Flexible Object-based Media Production

Davide Berghi, Craig Cieciura, Farshad Einabadi, Maxine Glancy, and 7 more authors

In ICME, 2024

Bib HTML

@inproceedings{berghi2024forecasterflexobm,
  title = {ForecasterFlexOBM: A Multi-View Audio-Visual Dataset for Flexible Object-based Media Production},
  author = {Berghi, Davide and Cieciura, Craig and Einabadi, Farshad and Glancy, Maxine and Camilleri, Oliver C and Foster, Philip and Nadeem, Asmar and Sardari, Faegheh and Zhao, Jinzheng and Volino, Marco and others},
  booktitle = {ICME},
  pages = {1--6},
  year = {2024},
  organization = {IEEE}
}

PAT

PAT: Position-Aware Transformer for Dense Multi-Label Action Detection

Faegheh Sardari, Armin Mustafa, Philip JB Jackson, and Adrian Hilton

In ICCV workshop, 2023

arXiv Bib HTML

@inproceedings{sardari2023pat,
  title = {PAT: Position-Aware Transformer for Dense Multi-Label Action Detection},
  author = {Sardari, Faegheh and Mustafa, Armin and Jackson, Philip JB and Hilton, Adrian},
  booktitle = {ICCV workshop},
  pages = {2988--2997},
  year = {2023}
}

VI-Pose-2

Unsupervised View-Invariant Human Posture Representation

Faegheh Sardari, Björn Ommer, and Majid Mirmehdi

In BMVC, 2021

arXiv Bib HTML

@inproceedings{sardari2021unsupervised,
  title = {Unsupervised View-Invariant Human Posture Representation},
  author = {Sardari, Faegheh and Ommer, Bj{\"o}rn and Mirmehdi, Majid},
  booktitle = {BMVC},
  year = {2021}
}

VI-Net

VI-Net: View-Invariant Quality of Human Movement Assessment

Faegheh Sardari, Adeline Paiement, Sion Hannuna, and Majid Mirmehdi

Sensors, 2020

arXiv Bib HTML

@article{sardari2020vi,
  title = {VI-Net: View-Invariant Quality of Human Movement Assessment},
  author = {Sardari, Faegheh and Paiement, Adeline and Hannuna, Sion and Mirmehdi, Majid},
  journal = {Sensors},
  volume = {20},
  number = {18},
  pages = {5258},
  year = {2020},
  publisher = {MDPI}
}

VI-Pose-1

View-Invariant Pose Analysis for Human Movement Assessment from RGB Data

Faegheh Sardari, Adeline Paiement, and Majid Mirmehdi

In ICIAP, 2019

Bib HTML

@inproceedings{sardari2019view,
  title = {View-Invariant Pose Analysis for Human Movement Assessment from RGB Data},
  author = {Sardari, Faegheh and Paiement, Adeline and Mirmehdi, Majid},
  booktitle = {ICIAP},
  pages = {237--248},
  year = {2019},
  organization = {Springer}
}

MGBSA-2

A Hybrid Occlusion Free Object Tracking Method Using Particle Filter and Modified Galaxy based Search Meta-Heuristic Algorithm

Faegheh Sardari, and Mohsen Ebrahimi Moghaddam

Applied Soft Computing, 2017

Bib HTML

@article{sardari2017hybrid,
  title = {A Hybrid Occlusion Free Object Tracking Method Using Particle Filter and Modified Galaxy based Search Meta-Heuristic Algorithm},
  author = {Sardari, Faegheh and Moghaddam, Mohsen Ebrahimi},
  journal = {Applied Soft Computing},
  volume = {50},
  pages = {280--299},
  year = {2017},
  publisher = {Elsevier}
}

MGBSA-1

An Object Tracking Method Using Modified Galaxy-based Search Algorithm

Faegheh Sardari, and Mohsen Ebrahimi Moghaddam

Swarm and Evolutionary Computation, 2016

Bib HTML

@article{sardari2016object,
  title = {An Object Tracking Method Using Modified Galaxy-based Search Algorithm},
  author = {Sardari, Faegheh and Moghaddam, Mohsen Ebrahimi},
  journal = {Swarm and Evolutionary Computation},
  volume = {30},
  pages = {27--38},
  year = {2016},
  publisher = {Elsevier}
}