In our paper, the final version of which was published this week in @NatureComms, we study the clinical task of switching patients from intravenous to oral antibiotic treatment. Paper: nature.com/articles/s41467-0… Code: github.com/SAFEHR-data/ivos-… 1/

Huge congratulations to my amazing PhD student, Niels D. Cariou Kotlarek, for leading this! In this work, we address a fundamental gap in generative TPPs: the lack of global sequence-level loss functions and mathematically justified distributional discrepancy measures. 2/

Very proud to be leading the SOFAIR Lab ucl.ac.uk/engineering/sofair… with fantastic colleagues from UCL, Edinburgh, Oxford and Cambridge.

📢 CFP: AgentSearch@SIGIR 2026 Submit your work on indexing, retrieval, and ranking of AI agents & tools. 📄 Extended abstracts (2p) & papers (4–9p) 📅 Deadline: Apr 15, 2026 (AoE) 🔗 agent-search.github.io/agent… #SIGIR #InformationRetrieval #AIagents

Our time series forecasting model “Sonnet” will be presented tomorrow by the amazing @yuxuan_shu at AAAI ‘26 (Singapore). 🗓️ Presentation details (Jan. 24) - Oral: 9:30–10:30am | Garnet 217 (ML: Time-Series/Data Streams 3) - Poster: afternoon #AAAI #AAAI2026 /1

I am offering three PhD projects at @ai_ucl / @uclcs as a part of the EPSRC Centre for Doctoral Training in Cyber-Physical Risk. More information about the CDT here: ucl.ac.uk/engineering/securi…. The projects are the following: - “Attack and Defence of Cyber-physical Systems Relying on Multimodal Foundational Models”; - “Hybrid Risks in Agentic AI Ecosystems”; - “Securing Cyber-physical Systems Against Cyber-attacks: A Hybrid Network Modelling Approach”. You can find more information about the projects at this link: ucl.ac.uk/engineering/securi…. Happy to arrange an informal chat if you are interested in applying.

New paper in Transactions on Machine Learning Research (@TmlrOrg) with @Gionceschelli: "DiffSampling: Enhancing Diversity and Accuracy in Neural Text Generation". Paper+code+video here: tmlr.infinite-conf.org/paper… @uclcs @ai_ucl

During the past year, our lab (in particular, @Gionceschelli and myself) have been involved in the @OECD Working Group on AI and the Future of Skills. The current goal of the Working Group is to devise a series of metrics to quantify AI capabilities on a scale from 1 to 5 (similar to the “autonomy” scale used for assessing the capabilities of self-driving cars) in order to derive a global standard for governments, businesses, and other organisations. In essence, the question is how to assess the extent to which an AI system can perform a given task with a level of proficiency comparable to that of humans. We focused on defining AI creative capabilities - not only in the artistic sense but, perhaps more importantly, in terms of problem-solving and original thinking. This work is based on @Gionceschelli 's PhD thesis and current projects. The OECD AI Capability Indicators Technical Report has now been published and is available here: oecd.org/en/publications/oec… The overall website about the OECD AI Capability Indicators project and policy work can be found here: aicapabilityindicators.oecd.…

I will be presenting our work on reward model overoptimisation in iterated RLHF at ARLET @ NeurIPS this afternoon. Work done with @_robertkirk and @mircomusolesi Paper: arxiv.org/abs/2505.18126 @FAICDT1 @uclcs @ai_ucl

Delighted to share that @yuxuan_shu's research has been accepted by #AAAI2026 for an oral presentation. Title: Sonnet: Spectral Operator Neural Network for Multivariable Time Series Forecasting Preprint: arxiv.org/abs/2505.15312 Repo: github.com/ClaudiaShu/Sonnet /1