Alejandro Tlaie Boria

Projects

Talks

Timeline

  1. 2025 to now

    AI oversight and verification

  2. Sep 2026

    Embedded Assessments for Frontier AI

    arXiv With Jacob Charnock, Sophie Williams, Zaheed Kara and 4 others

  3. Sep 2026

    How to scale the impact of embedded AI evaluators

    OpenMined blog With Andrew Trask, Jack Hardinges, Koen van der Veen and Lacey Strahm

  4. Jul 2026

    Confidential computing can enable better frontier AI auditing

    Pour Demain technical brief

  5. Jun 2026

    Expanding External Access to Frontier AI Models for Dangerous Capability Evaluations

    ACM FAccT 2026 With Jacob Charnock, Kyle O’Brien, Stephen Casper and Aidan Homewood

  6. May 2026

    Securing deep frontier AI evaluations

    Pour Demain technical brief

    Poster version, “Securing deep third-party frontier AI evaluations”, at the ICML 2026 Second Workshop on Technical AI Governance Research.

  7. 2025

    A Blueprint for an EU Ecosystem of Secure, Deep and External AI Audits

    SSRN

    Poster version, “A Blueprint for a Secure EU AI Audit Ecosystem”, at the ICML 2025 Workshop on Technical AI Governance (TAIG).

  8. Mar 2025

    Securing External Deeper-than-black-box GPAI Evaluations

    arXiv With Jimmy Farrell

  9. Apr to Oct 2025

    AI risk quantification

  10. Dec 2025

    Technical report: LLM-simulated expert judgement for quantitative AI risk estimation

    SaferAI With Otter Quarks, Jakub Kryś, Siméon Campos and Henry Papadatos

  11. Dec 2025

    A Methodology for Quantitative AI Risk Modeling

    arXiv With Malcolm Murray, Steve Barrett, Henry Papadatos and 4 others

  12. Dec 2025

    Toward Quantitative Modeling of Cybersecurity Risks Due to AI Misuse

    arXiv With Steve Barrett, Malcolm Murray, Otter Quarks and 16 others

  13. 2024 to 2025

    AI safety research

  14. Feb 2025

    Assessing confidence in frontier AI safety cases

    arXiv With Stephen Barrett, Philip Fox, Joshua Krook, Tuneer Mondal and Simon Mylius

  15. Dec 2024

    delphi-llms

    GitHub

    An LLM-based Delphi expert elicitation pipeline.

  16. Oct 2024

    A short project on Mamba: grokking & interpretability

    LessWrong

  17. Oct 2024

    Using AI Alignment Theory to understand the potential pitfalls of regulatory frameworks

    arXiv

  18. Oct 2024

    Unlearning safety fine-tuning through SAE latent steering

    ARENA capstone project, GitHub With AnimalCracker and simonwisdom

  19. Aug 2024

    Toy Models of Superposition: what about BitNets?

    LessWrong

  20. Jun 2024

    Introducing SARA: a new activation steering technique

    LessWrong

  21. 2025

    Exploring and Steering the Moral Compass of Large Language Models

    ICPR 2024 workshops, Springer LNCS 15619

    Ethical dilemmas, moral values profiles, a novel causal intervention...

    This study has it all!

    Two network diagrams of eight schools of ethics, among them virtue ethics, deontology, and act and rule utilitarianism: one for proprietary language models and one for open models. Coloured arrows between the schools show transition probabilities from 0 to 1.
  22. 2019 to 2024

    Neuroscience and information theory

  23. Jun 2025

    Inferring internal states across mice and monkeys using facial features

    Nature Communications With Muad Y. Abd El Hay, Berkutay Mert, Robert Taylor and 6 others

    Used a data-driven approach (GLM-HMM) to predict an animal’s performance in a naturalistic setup by only having access to facial expression.

    Conference poster, “Thoughtful faces: Using facial features to infer naturalistic cognitive processing across species”, showing the virtual-reality task for macaques and mice, their task performance, a GLM-HMM model that infers internal states from facial features, and the conclusions
  24. May 2025

    A Python toolbox for neural circuit parameter inference

    npj Systems Biology and Applications With Alejandro Orozco Valero, Víctor Rodríguez-González, Noemi Montobbio and 6 others

  25. Apr 2025

    Diverging roles of TRPV1 and TRPM2 in warm-temperature detection

    eLife With Muad Y Abd El Hay, Gretel B Kamm and Jan Siemens

  26. Apr 2024

    What does the mean mean? A simple test for neuroscience

    PLOS Computational Biology With Katharine Shapcott, Thijs L. van der Plas, James Rowland and 6 others

    Developed a general and simple computational method that allows neuroscientists to check the relevance of trial-averages (such as firing rates, PSTHs) in their datasets. Validated this method in two different datasets.

    Conference poster, “Does the brain average?”, testing whether trial-averaged neural responses are stimulus-specific and useful for behaviour in two mouse datasets, one with direct neuronal stimulation and one with sensory stimulation
  27. 2016 to 2020

    Complex networks

  28. 2020

    PhD Thesis in Complex Networks

    Technical University of Madrid

    Relating structure and dynamics of nonlinear systems (chaotic and stochastic) that are embedded in complex networks; extra: experimental validation.

    On top of that, phenomenological model of the neuro-glial interaction in Spiking Neural Networks!

    Theoretical prediction beside experimental result: a model network and a culture of neurons, each with plots of complexity against coupling strength, where low-degree nodes and hubs behave differently, and of complexity falling as node degree grows
  29. Nov 2019

    High-order couplings in geometric complex networks of neurons

    Physical Review E With I. Leyva and I. Sendiña-Nadal

  30. Theoretical physics

  31. MSc Thesis in String Theory

    Complutense University of Madrid

    Spatial and temporal solution for the pulsating string in an AdS³ ambient space. These results generalize and extend previously known ones. Furthermore, we can explain them in a more parsimonious way.

    Excerpt from the MSc thesis: equations 3.14 to 3.16, which give the final solution for the pulsating string in terms of Jacobi elliptic functions

Side projects