Last week, our PhD students and academic staff traded classrooms and computer screens for countryside views and conversations as we embarked on a three-day retreat designed to recharge and reconnect. The retreat offered a welcome break from daily routines and delivered a memorable mix of insightful talks, friendly competition, reflective workshops—and some great food.

Here’s a look back at the highlights.

The retreat kicked off with a warm welcome and a delicious lunch that set the tone for what would be an engaging and collaborative few days. With everyone gathered in one place, our first structured activity was a team-building session designed to help us connect beyond the confines of our usual roles. Through creative problem-solving and small-group activities, we challenged ourselves to think outside of the box, communicate more effectively, and collaborate in ways that pushed us beyond our comfort zones!

Later in the afternoon, we transitioned into a series of student-led talks. PhD students presented snapshots of their research, prompting engaging discussions and questions.  

The day wrapped up with a fabulous dinner and an informal but highly entertaining evening comprising of impromptu chat and friendly rivalry over board games, breaking the ice even further and setting a welcoming tone for the retreat. 

We began Day 2 with another round of student talks. Building on the previous day, the presentations were dynamic, insightful, and spurred plenty of discussion. The talks showcased the incredible breadth of work happening within our centre and it was  inspiring to see the students speak about their work.

Lunch gave us another chance to unwind and continue conversations sparked during the talks. Then, it was time for the second team-building activity—a treasure hunt around the local village that quickly became a highlight of the retreat.

Armed with clue sheets, maps, and a healthy sense of competition, teams set off on foot to explore the village, solving riddles and deciphering hints that led them to landmarks, a cemetery (!) and hidden details. The experience was packed with laughter, a bit of friendly competition, and plenty of teamwork. Regardless of who “won,” the real reward was the sense of connection and discovery that came with each solved clue.

Back at the retreat venue, dinner that evening felt like a celebration of everything we’d already achieved together. Conversations stretched into the night, blending personal stories with professional ideas. Whether discussing research struggles or favourite hobbies, everyone was enjoying the rare chance to connect informally and meaningfully.

Our final day featured something a little different: a Lego Serious Play workshop – a guided session using Lego bricks to express ideas, metaphors, and visions for our academic and personal journeys. This hands-on session invited participants to use Lego bricks to build metaphors and models that reflected their academic experiences, personal growth, and future aspirations. What might sound whimsical on the surface turned out to be deeply reflective and was a powerful way to close the retreat: creative, thoughtful, and uniquely personal.

Following a final lunch together, we said our goodbyes—refreshed, inspired, and with a renewed appreciation for the people who make up our academic community.

The retreat was more than just a break from routine. It was a space for conversation, connection, and creativity. We came away with new ideas, stronger relationships, and a lego pack! To everyone who made it happen—thank you. Here’s to more moments like these in the future.

I recently had the pleasure of presenting our work “Private Selection with Heterogeneous Sensitivities” at the 3rd IEEE Conference on Secure and Trustworthy Machine Learning (SaTML2025) in Copenhagen. This work was presented in the form of a poster as well as a talk at the main conference.

• We’re working on further improving the adaptive mechanism — several exciting open questions remain.
• We’re also exploring the connection to online learning, especially in settings with distribution shift. We only hinted at this in the paper, but early results are promising and suggest real benefits from private selection in sequential decision-making.

I attended the International Conference on 3D Vision (3DV) in March. 3DV is a conference that brings together researchers in 3D computer vision and graphics. At the conference, I presented our paper, Spann3R: 3D Reconstruction with Spatial Memory, which was accepted by 3DV as an award candidate.

Our work addresses the problem of dense 3D reconstruction from images only, without any prior knowledge of camera parameters or depth information. Due to the inherent ambiguity of interpreting 3D structures from 2D images, traditional dense reconstruction pipelines typically decompose the task into a series of minimal problems, which are then chained together. This approach usually requires substantial manual tweaks as well as engineering efforts. In contrast, our approach directly maps a set of 2D images to 3D using a neural network, enabling a fully feed-forward dense reconstruction.

The key idea of our method is to manage a spatial memory that stores previous states and learns to query this spatial memory to reconstruct the geometry of the subsequent frames. Our results demonstrate a proof-of-concept for feed-forward 3D reconstruction, and open up exciting directions for fully learning-based approaches to Structure-from-Motion (SfM) and Simultaneous Localization and Mapping (SLAM). Interestingly, as a by-product, our method can perform dense reconstruction without explicitly estimating camera poses—a concept that some later works have referred to as pose-free reconstruction.

Beyond my own presentation, I attended several insightful talks, including those by Jon Barron, Noah Snavely, and Fei-Fei Li. In Jon’s talk, he mentioned one particular interesting topic on “Why care about 3D?”. He discussed various perspectives on the role of 3D in computer vision and robotics.

Jon challenged the traditional assumption that a robot must reconstruct its environment in 3D to navigate it effectively. With the progress of deep learning, robots can map observations directly to control signals without explicit 3D modeling—similar to trends in autonomous driving. Likewise, the idea that realistic image or video generation requires full 3D reconstruction is challenged with the rapid progress of video diffusion models.

Jon offered several interesting arguments for why 3D still matters. One reason is efficiency—once a 3D scene is reconstructed, it can be rendered from many viewpoints at low cost. More fundamentally, humans live in a 3D world—so if we expect AI systems to perceive and interact like us, they must reason in 3D too.

Yet, the community has not reached a consensus on how to let AI models learn 3D efficiently. While vast amounts of 2D image data are readily available thanks to smartphones and the Internet, acquiring large-scale, high-quality 3D data remains costly. Hopefully, the progress in 3D foundation models can unlock new ways to scale up the learning of 3D representations from web-scale casual videos without ground-truth 3D, and bridge the gap between 2D and 3D through unified multi-modal representations.

Further information on Hengyi’s project can be found here https://hengyiwang.github.io/projects/spanner

Last month, CDT students travelled to Vancouver to present their work at NeurIPS, one of the largest AI conferences. The schedule was packed, with 6 conference papers and 3 workshop papers presented. The CDT is proud to sponsor and support our PhD students as they present their research at the conference, showcasing their hard work and academic excellence on an international stage!
Day 1, Tuesday: No papers were presented, but the expo, tutorials, and careers fair kept everyone occupied.

Day 2, Wednesday:

William Bankes William and his supervisor presented their work on robust data down sampling. Naive approaches to training on a subset of data can cause problems when classes in the dataset are imbalanced, with rare classes becoming even rarer. A direct result of this is the model’s ability to predict on this subset of the data decreases. To address this they proposed an algorithm called REDUCR, which downsamples data in a manner that preserves the performance of minority classes within the dataset. They show REDUCR works across a range of text and image problems achieving state of the art results. The work is available here.

Jake Cunningham also presented his work on 
Reparameterized Multi-Resolution Convolutions for Long Sequence Modelling

Day 3, Thursday
Yuchen Zhu 

Yuchen’s joint paper with Jialin Yu and Ricardo Silva from UCL Statistical Sciences, Structured Learning of CompositionalSequential Models was presented as part of the main proceedings of NeurIPS 2024. They proposed an explicit model for expressing how the effect of sequential interventions can be isolated into modules, clarifying previously unclear data conditions that allow for the identification of their combined effect at different units and time steps. The paper is here.  Additionally, together with collaborators from MPI Tuebingen, Yuchen presented a paper Unsupervised Causal Abstraction at the Causal Representation Learning workshop. Due to the lack of interpretability of current large blackbox models, they propose a methodology for causally abstracting a large model to a smaller and more interpretable model. In particular, unlike existing methods, their method does not require supervision signals from the smaller model. The paper can be found  here.

David Chanin

David presented a paper along with co-author Daniel Tan, another UCL PhD Student. They find that Contrastive Activation Addition (CAA) steering has mixed results in terms of robustness and reliability. Steering vectors tend to generalise out of distribution when they work in distribution. However, steerability is highly variable across different inputs: depending on the concept, spurious biases can substantially contribute to how effective steering is for each input, presenting a challenge for the widespread use of steering vectors. While CAA is effective on some tasks, other behaviours turn out to be unsteerable. As a result, it is difficult to ensure they will be effective on a given task of interest, limiting their reliability as a general alignment intervention. The paper is available here.

Day 4, Friday
Reuben Adams

Reuben presented his paper (co-authored with his supervisors) on extending a classic theorem in the PAC-Bayes literature to account for arbitrary outcomes, rather than simply correct or incorrect classifications. Their work provides theoretical insight into the generalisation behaviour of neural networks and the different kinds of errors they can make. Their framework can cover not just Type I and Type II errors, but any kind of error that may occur in multi class classification. You can find the paper here.

Daniel Augusto
Daniel presented a co-authored paper in collaboration with
Getúlio Vargas Foundation for the main conference track. Their work proposes a new solution to streaming variational Bayesian inference using  GFlowNets as a foundation for their methodology. This was the first work that allows high quality variational inference for discrete parameters without requiring the storage of the whole dataset. They believe this work will be useful for applications in genetics, through the inference of phylogenetic trees, for preference learning, and other big-data contexts. Their paper can be read here.

Day 5, Saturday 

Oscar Key Oscar, along with co-authors from Graphcore presented a poster at the workshop on Efficient Natural Language and Speech Processing. Their work considers the top-k operation, which finds the largest k items in a list, and investigates how it can be computed as quickly as possible on the parallel hardware commonly used to run AI applications. Top-k is found in many AI algorithms, so it’s useful to make it fast, e.g. a large language model might use it to select the most important parts of a long prompt. Their full paper is available here.

Varsha Ramineni 

Varsha presented her research at the Workshop on Algorithmic Fairness through the Lens of Metrics and Evaluation (AFME 2024). This work addresses the challenge of evaluating classifier fairness when complete datasets, including protected attributes, are inaccessible. They propose a novel approach that leverages separate overlapping datasets, such as internal datasets lacking demographic information  and external sources like census data, to construct synthetic test data with all necessary variables. Experiments demonstrate that our approach produces synthetic data with high fidelity and offers reliable fairness evaluation where real data is limited.  Varsha says that she had an incredible experience attending her first NeurIPS and presenting her work and engaging in meaningful discussions throughout the conference was a deeply rewarding experience, providing invaluable feedback and ideas as the work extends. Do reach out to her if you’d like to learn more!

My research revolves around kernel-based Monte Carlo methods and generative models—fields I’m eager to see advance with both solid theoretical underpinnings and real-world impact. I was delighted to accept an invitation from Professor Chris Oates to visit School of Mathematics, Statistics and Physics of Newcastle University and attend a workshop on extrapolation methods as part of the Visiting Fellowships in Machine Learning. This visit was a fantastic chance to share my work, learn from the excellent researchers there, and discuss new ideas.

Beyond the academic environment, I was struck by the warmth of Newcastle as a city, with its stunning Quayside views and friendly local community. I’m grateful to Professor Oates and the entire team from the School of Mathematics, Statistics and Physics, Newcastle University for hosting me. I really look forward to future opportunities for more collaboration on research topics.

2019 Newcastle-upon-Tyne University Open Day, Tyne and Wear, England, UK. This is the logo for the University on the side of one of the buildings in the public grounds of the University campus.

As I return to my PhD studies at UCL after a transformative six-month internship at Amazon Robotics in Berlin, I find myself reflecting on the invaluable experiences and lessons gained during this period. I am eager to share insights from my journey and how they have shaped my thinking.
During my time at Amazon, I was immersed in a dynamic environment where innovation and collaboration were at the forefront. My role involved developing computer vision algorithms leveraging robot-object interactions. The challenges were both stimulating and rewarding. Working alongside brilliant colleagues, I witnessed first-hand the power of collective intelligence in tackling complex problems. The startup-like atmosphere fostered a culture of creativity and agility, enabling us to iterate and adapt swiftly to evolving demands.

This internship experience provided me with a deeper understanding of the importance of leadership principles in driving organizational success. The team’s well-defined values served as guiding beacons, shaping not only our professional conduct but also our approach to problemsolving. These principles, characterized by a customer obsession, working backwards from the goal, and relentless innovation, are invaluable assets that I intend to carry forward in my PhD and future career endeavors.
One of the most significant takeaways from my internship is the cultivation of an entrepreneurial mindset. Encouraged to think outside the box and embrace experimentation, I learned to approach challenges with a blend of creativity and pragmatism. This adaptive mindset enabled me to navigate uncertainty and also empowered me to explore new avenues for growth and development.
Would I recommend doing an internship at Amazon? Absolutely.
As I transition back to academia, I am grateful for the rich experiences and invaluable lessons learned during my internship, and for the opportunity to contribute to cutting-edge robotics. Armed with newfound insights and skills, I am eager to apply them to my research at UCL and beyond. I am confident that the lessons learned and connections made during my time in Berlin will continue to influence and inspire my professional journey.

PS. For more information on how Amazon is approaching robotics and robot learning, check out https://www.aboutamazon.com/news/operations/how-amazon-deploys-robots-in-its-operations-facilities

The UKRI Centre for Doctoral Training in Foundational AI is proud to announce the participation of Sierra Bonilla in the International Computer Vision Summer School (ICVSS) 2024, held in Sicily from July 7 to July 13, 2024. This event is known for its rigorous selection process, with 30% of applicants accepted from a pool of 626.

This year, the theme focused on “Computer Vision in the Age of Large Language Models,” covering cutting-edge topics at the forefront of research and industry innovation. The summer school featured 30 hours of intensive lectures delivered by world-renowned experts, including:

Vijay Badrinarayanan (Wayve, USA)

Michael J. Black (Max Planck Institute for Intelligent Systems and Meshcapade GmbH, DEU)

Andreas Geiger (University of Tübingen, DEU)

Georgia Gkioxari (California Institute of Technology, USA)

Abhishek Gupta (University of Washington, USA)

Aaron Hertzmann (Adobe Research, USA)

Derek Hoiem (University of Illinois at Urbana-Champaign, USA)

Diane Larlus (Naver Labs Europe, FRA)

Richard Newcombe (Meta Reality Labs Research, USA)

Pietro Perona (California Institute of Technology, USA)

Bobby Rao (Microsoft, USA)

Stefano Soatto (Amazon and University of California Los Angeles, USA)

Andrea Tagliasacchi (Simon Fraser University, CAN)

Antonio Torralba (Massachusetts Institute of Technology, USA)

Lior Wolf (Tel Aviv University & Mentee Robotics, ISR)

One of the highlights of the event was the interactive poster sessions, where Sierra presented her recently accepted MICCAI ’24 Paper “Gaussian Pancakes”. The social events and the beautiful Sicilian environment and food made networking with peers and experts very memorable.

Special thanks go to the ICVSS 2024 organizers, particularly Giovanni Maria Farinella, Roberto Cipolla, and Sebastiano Battiato, for orchestrating such an incredible event. Additional gratitude is extended to Derek Hoiem for his valuable mentorship session.

The knowledge and insights gained from ICVSS 2024 are expected to greatly benefit Sierra Bonilla’s ongoing research and future projects, further contributing to the advancement of the field within the UKRI CDT in Foundational AI.

On Wednesday 15th November, members of all cohorts within the FAICDT visited our Robotics labs at the newly opened UCL East Campus. Prof Dimitrios Kanoulas very kindly offered up his time and expertise in showing our students around both the Marshgate and Pool Street Labs.

Our students were able to interact with the quadra-ped robots and speak with academics and other researchers about their ground breaking work in the field of AI and robotics.

The visit to UCL East’s robotics lab was impressive. We saw the Boston Dynamics robot navigate stairs and avoid obstacles, showcasing the practical applications of these technologies. The most striking moment was seeing a robot execute a backflip, which highlighted the advanced capabilities in robotics. Dimitrios Kanoulas kindly also gave us a tour around the Marshgate building! The location’s calm park setting was a nice change from the main campus environment. – Sierra Bonilla, Cohort 5

A great time was had by all, and we are delighted that our students have been able to connect with other researchers in the AI field here at UCL.

For the second year running, the UKRI CDT in Foundational AI, collaborated with the University of Bath and the University of Bristol to organise an inter CDT conference at the Bristol Hotel.

This year our CDT put on a session on AI Vision, organised with Prof Lourdes de Agapito, and a student-led session on the Threat of AI, organised with Reuben Adams and Robert Kirk.

The event began with a welcome to all from our FAICDT Deputy Director Gabriel Brostow, followed on by a Keynote Speech by Jasmine Grimsley and Sarah-Jane Smyth from the London Data Company.

The AI Vision session opened up the individual sessions conference on day one, and we were delighted to host three eminent academics to discuss their work and research. First we had Christian Rupprecht from the University of Oxford discussing ‘Unsupervised Computer Vision in the Time of Large Models’. He was followed on by Laura Sevilla-Lara from the University of Edinburgh, who discussed her work on ‘Efficient Video Understanding’ and the session finished with Edward Johns from Imperial College London, discussing his research on ‘Vision-Based Robot Learning of Everyday Tasks’. The room was filled with an eager audience who asked multiple questions of each speaker, with a real interest and excitement on the work that was being undertaken by all in this field.

The rest of the first day saw sessions from the other two CDTs involved, with an agenda for the event available here: UKRI Inter AI CDT Conference 2023 – ART-AI (cdt-art-ai.ac.uk)

As with our first conference, there was a popular poster session held at the MShed in the afternoon of Day 1, with prizes handed out at our evening dinner. Our new, cohort 5, student Ahmet Guzel was the winner from the FAICDT with a poster that won outstanding marks from all judges on the day.

Day two started with a keynote speech from our final year student Jakob Zeitler, who is currently on interruption whilst working on his start up company Matterhorn Studios. Jakob gave an insightful talk on ‘Machine Learning for Material Science: Using Bayesian Optimisation to Create a Sustainable Materials Future’. We were delighted to have Jakob provide a keynote here and it was wonderful to see so many individuals reach out to him after his talk to find out more about his research and work at Matterhorn.

The afternoon session by the FAICDT was led by students Reuben Adams and Robert Kirk. This was a vibrant session, asking attendees to think about where they currently sat on the spectrum of concern with regards to the safety of AI and then have in depth discussions with one another about possible AI solutions and potential regulations and possibilities for future safety. The event was interactive, and highly engaging!

The event closed with a final keynote by Steven Schokaert from Cardiff University with a networking session for all students before returning to their home institutions.

All in all, another successful event! With thanks to Brent Kiernan, Christina Squire and Suzanne Binding for being co-organisers extraordinaire!

I am interested in discussing an intriguing concept in machine learning, which promises to revolutionize the way we approach learning in robotics: World Models.

At a high level, World Models aim to create a compact and controllable representation of the world. Think of it as a mental simulation or an internal mini-world where AI can experiment, explore, and ‘imagine’ different scenarios, all without the need for real-world interactions. It’s like creating a sandbox game for AI, where it can learn the ropes before stepping out into the real world. ??

Let’s contrast this with the conventional end-to-end learning methods. These traditional approaches typically require vast amounts of real-world data and intensive training, which can be time-consuming, computationally expensive, and let’s face it, data-inefficient.

That’s where the beauty of World Models shines. By allowing AI to ‘dream’ or simulate possible scenarios in their internal model of the world, they can learn faster and more efficiently. They can plan and strategize better by running various ‘what-if’ scenarios within their world model. Imagine playing chess and being able to simulate all possible moves in your mind before making your move – that’s the advantage of World Models in a nutshell! ??

The ‘DayDreamer’ paper is a fantastic resource if you’re keen to delve into the specifics of this innovative approach. It opens up new vistas in our quest for smarter and more data-efficient learning in robotics.

In a world where data is king but also a constraint, World Models are pioneering a path towards more strategic, efficient, and thoughtful AI. So, let’s continue learning, exploring, and innovating. After all, the future of AI is as exciting as we dare to imagine!

#WorldModels #MachineLearning #DataEfficiency