About Me

I am a final-year ELLIS PhD student at UCL AI Centre, working on Causality and LLM Alignment, fortunately supervised by Matt Kusner, Ricardo Silva and Dominik Janzing(ELLIS). I am also grateful to have worked extensively with Arthur Gretton and for the wonderful times at Amazon Research Tuebingen and Microsoft Research Cambridge. In a previous life, I studied Maths at Cambridge.

My research interests are three-fold: 1. understanding the foundations of how causal structures/effects arise from fine-grained models (CLeaR 2024, NeurIPS CRL 2024), 2. developing effect estimation algorithms with guarantees (ICML 2021, UAI 2022, NeurIPS 2021, NeurIPS 2024), and 3. the foundations of the LLM reward hacking problem and how it can be solved (ICML 2025).

News

05.2025 New work: When Can Proxies Improve the Sample Complexity of Preference Learning? has been accepted to ICML 2025.

03.2024 I gave a talk on causal abstraction and the paradoxes which arise from it at Valence CARE reading group. The talk is recorded, check it out if you want to learn about the fundamentals and consequences of causal abstraction!

01.2024 New work: Meaningful Causal Aggregation and Paradoxical Confounding has been accepted to CLeaR 2024.

Older news

04.2023 I gave an invited talk on Causal Inference Under Treatment Measurement Error: A Nonparametric Instrumental Variable Approach at Causality Seminar (China). Talk slides are here.

04.2023 I gave an invited talk on confounded causal inference with proxies or measurement error at Causal Inference Meets Statistics Quarterly Meeting. Talk slides are here and the poster is here.

01.2023 I am thrilled to join Microsoft Research Cambridge to work with Cheng Zhang on causal inference with many measurements from 01.2023 to 03.2023.

06.2022 I am thrilled to join Amazon Science Tuebingen to work with Dominik Janzing on abstracting causal models from 06.2022 to 12.2022.

05.2022 New work: Causal Inference Under Treatment Measurement Error: A Nonparametric Instrumental Variable Approach has been invited for an oral presentation at UAI 2022.

05.2022 I am presenting my work on causal effect estimation with latent variables(slides) at the Statistics for Data-Centric Engineering Seminar Series, Alan Turing Institute and on structured treatment effect estimation(slides) at Professor Menggang Yu’s group, University of Wisconsin, Madison.

10.2021 New work: Causal Effect Estimation for Structured Treatments has been accepted to NeurIPS 2021.

05.2021 New work: Proximal Causal Learning with Kernels: Two-Stage Estimation and Moment Restrictions has been accepted to ICML 2021.