Nino Scherrer Portrait

Nino Scherrer

nino.scherrer[at]gmail.com

I am a research scientist at Patronus AI working on automated language model evaluations. Apart from that, I am heavily interested intested in causality, cognitive and social science, and all forms of synthetic data. I love meeting new people through research collaborations - if you are interested in working with me, please get in touch!

Previously, ...

Fields of Interest

Rigorous LLM Evaluations Causality AI Safety / AI Ethics Interpretability Model Robustness

Research

> Language Model Evaluations

> Mechanistic Interpretability

  • Uncovering Mesa-Optimization Algorithms in Transformers
    Johannes von Oswald*, Eyvind Niklasson*, Maximilian Schlegel*, Seijin Kobayashi, Nicolas Zuccet, Nino Scherrer, Nolan Miller, Mark Sandler, Blaise Agüera y Arcas, Razvan Pascanu and João Sacramento
    Arxiv Preprint, 2023

    Paper
    Mechanistic Interpretability
    Mesa-Optimization

> Causality / Machine Learning

> Medical Image Analysis

> Biology / Tissue Engineering

Tutorial "Causality and Deep Learning"

I contributed to the ICML 2022 Tutorial on "Causality and Deep Learning: Synergies, Challenges and the Future". The corresponding work (systematic survey) "Deep Learning for Causality: A Unifying Perspective on Neural Causal Structure Learning" will be released soon.
Tutorial Slides

Talks

Upcoming:
  • May 2024: "Evaluating (Moral) Beliefs Encoded in LLMs", MilaNLP Lab @ Bocconi University, Milan
Past:
  • September 2023: "Evaluating the Moral Beliefs Encoded in LLMs", ChatGPT Zurich Group (AI Meetup), Zurich
  • February 2023: "Deep Learning for Causality: Neural Causal Structure Learning", AI for Actionable Impact (AI4AI) Lab , Imperial College London
  • January 2023: "On the Synergies of Causality and DeepLearning", Neuroscience in ML Seminar, ETH Zurich
  • April 2022: "Learning Neural Causal Models with Active Interventions", XAI seminar series, Imperial College London
  • October 2021: "Learning Neural Causal Models with Active Interventions", Causality Reading Group, TU Darmstadt

Nino's Log

Heavily inspired by Liliam Weng's Log, I am starting to "log" my learning notes / thoughts in the field of AI using blog posts.




Last Updated: April, 2024