Mimicking User Data: On Mitigating Fine-Tuning Risks in Closed Large Language Models
F. Eiras, A. Petrov, P. H. S. Torr, M. Pawan Kumar, A. Bibi
Workshop on the Next Generation of AI Safety (NextGenAISafety), ICML, 2024
workshop
paper
Risks and Opportunities of Open-Source Generative AI
F. Eiras, A. Petrov, B. Vidgen, [...], T. Darrell, Y. Lee, J. Foerster
arXiv, 2024
arxiv
paper
Near to Mid-term Risks and Opportunities of Open-Source Generative AI
F. Eiras, A. Petrov, B. Vidgen, [...], T. Darrell, Y. Lee, J. Foerster
International Conference of Machine Learning (ICML) - Position Paper, 2024
conference
oral
paper
Efficient Error Certification of Physics-Informed Neural Networks
F. Eiras, A. Bibi, R. Bunel, K. Dj Dvijotham, P. H. S. Torr, M. Pawan Kumar
International Conference of Machine Learning (ICML), 2024
conference
paper
Segment, Select, Correct: A Framework for Weakly-Supervised Referring Segmentation
F. Eiras, Kemal Oksuz, A. Bibi, P. H. S. Torr, P. K. Dokania
arXiv, 2023
preprint
paper
Provably Correct Physics-Informed Neural Networks
F. Eiras, A. Bibi, R. Bunel, K. Dj Dvijotham, P. H. S. Torr, M. Pawan Kumar
2nd Workshop on Formal Verification of Machine Learning, ICML, 2023
workshop
outstanding paper award
paper
Towards Certified Machine Learning
F. Eiras
Presented at Columbia University (hosted by Prof. Suman Jana), New York University (hosted by Prof. Jinyang Li & Prof. Joseph Tassardi), University of California, Berkeley (hosted by Prof. Somayeh Sojoudi), and Stanford University (hosted by Prof. Clark Barret)
talk
slides
Certifying Ensembles: A General Certification Theory with S-Lipschitzness
A. Petrov*, F. Eiras, A. Sanyal, P. H. S. Torr, A. Bibi*
International Conference of Machine Learning (ICML), 2023
conference
paper
RANCER: Non-Axis Aligned Anisotropic Certification with Randomized Smoothing
T. Rumezhak, F. Eiras, P. H. S. Torr, A. Bibi
IEEE/CVF Winter Conference on Applications of Computer Vision (WACV), 2023
conference
paper
ANCER: Anisotropic Certification via Sample-wise Volume Maximization
F. Eiras*, M. Alfarra*, M. Pawan Kumar, P. H. S. Torr, P. K. Dokania, B. Ghanem, A. Bibi*
TMLR, 2022
journal
paper
A Two-Stage Optimization-based Motion Planner for Safe Urban Driving
F. Eiras, M. Hawasly, S. V. Albrecht, S. Ramamoorthy
IEEE Transactions on Robotics (T-RO), 2021
journal
paper bibTeX
PILOT: Efficient Planning by Imitation Learning and Optimisation for Safe Autonomous Driving
H. Pulver, F. Eiras, L. Carozza, M. Hawasly, S. V. Albrecht, S. Ramamoorthy
IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2021
conference
paper
Interpretable Goal-based Prediction and Planning for Autonomous Driving
S. V. Albrecht, C. Brewitt, J. Wilhelm, B. Gyevnar, F. Eiras, M. Dobre, S. Ramamoorthy
IEEE International Conference on Robotics and Automation (ICRA), 2021
conference
paper bibTeX
Correct-by-Construction Advanced Driver Assistance Systems based on a Cognitive Architecture
F. Eiras, M. Lahajinan, M. Kwiatkowska
IEEE Connected and Automated Vehicles Symposium (CAVS), 2019 (Oral presentation)
conference
paper bibTeX
Analytical Modeling of Vanishing Points and Curves in Catadioptric Cameras
P. Miraldo, F. Eiras, S. Ramalingam
IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2018
conference
paper bibTeX
* indicates equal contribution.
See all in Google Scholar.
experience
Five, Dec. 2022 - Jun. 2023
Research Scientist Intern
Worked on efficient methods to perform zero-shot and weakly-supervised referring image segmentation (i.e., segmenting an object in an image that is referred in a natural language sentence), achieving new state-of-the-art performance in the field.
Five, Jun. 2021 - Sep. 2021
Research Scientist Intern
Extended the certified robustness technique of randomized smoothing from isotropic isotropic balls to anisotropic certificates through a simplified Lipschitz analysis-based framework.
Five, Sep. 2018 - Sep. 2020
Research Engineer, Motion Planning and Prediction
• Led the development of safe and scalable optimization-based motion planning algorithms, working in a team with research scientists and software engineers
• Published and presented research work developed at top tier conferences and journals within the robotics community, as well as to non-technical audiences
• Wrote and reviewed research and development code, ensuring CI with other tools within the company
• Carried out literature reviews, prepared and presented cutting edge research at paper reading groups
Institute for Systems and Robotics, Lisbon, Apr. 2017 - Sep. 17
Graduate Research Assistant
Studied and developed new methods to perform pose estimation through vanishing points in general (central and non-central) omnidirectional cameras which lead to the publication of a paper in one of the top conferences in the field, IEEE CVPR 2018.
education
University of Oxford, Oct. 2020 - Oct. 2024 (Expected)
PhD in Engineering Science (AIMS)
Relevant courses: Data Estimation & Inference, Machine Learning, Reinforcement Learning, Discriminative & Deep Learning for Big Data
Supervisors: Prof. Philip H.S. Torr, Dr. Adel Bibi and Dr. Pawan Kumar (Google DeepMind)
University of Oxford, Oct. 2017 - Sep. 2018
MSc in Computer Science (Distinction)
Relevant courses: Computational Game Theory, Probabilistic Model Checking
Dissertation: “To Err is Human: Designing Correct-by-Construction Driver Assistance Systems using Cognitive Modelling”
Supervisors: Dr. Morteza Lahijanian and Prof. Marta Kwiatkowska
pdf code
EPFL, Sep. 2016 - Feb. 17
Student Exchance (5.75/6)
Relevant courses: Applied Machine Learning, Image and Video Processing, Lab in Image and Signal Processing
Awarded a monthly scholarship under the Swiss-European Mobility Programme
Técnico Lisboa, Sep. 2013 - Jul. 2016
BSc in Electrical and Computer Engineering (18/20)
Relevant courses: Algorithms and Data Structures, Signals and Systems, Computational Mathematics, Automatic Control
Top 2% of the class
Academic Excelency Award every year and for the end of the BSc (3 years)
contact
Francisco Eiras
Linacre College
St.Cross Rd,
Oxford OX1 3JA,

Check out my Google Scholar and Github, or say 👋 via Twitter, LinkedIn or email