Publications

CLAM: Selective Clarification for Ambiguous Questions with Large Language Models
Lorenz Kuhn, Yarin Gal, Sebastian Farquhar
ICML 2023 Workshop on Deployment Challenges for Generative AI, arXiv

Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation
Lorenz Kuhn, Yarin Gal, Sebastian Farquhar
ICLR2023 (Spotlight)

Robustness to Pruning Predicts Generalization in Deep Neural Networks
Lorenz Kuhn, Clare Lyle, Aidan N. Gomez, Jonas Rothfuss, Yarin Gal
arXiv

Efficient Smoothing of Dilated Convolutions for Image Segmentation
Thomas Ziegler, Manuel Fritsche, Lorenz Kuhn, Konstantin Donhauser
arXiv

Patient Risk Assessment and Warning Symptom Detection Using Deep Attention-Based Neural Networks
Ivan Girardi, Pengfei Ji, An-phi Nguyen, Nora Hollenstein, Adam Ivankay, Lorenz Kuhn, Chiara Marchiori, Ce Zhang
Health Text Mining and Information Analysis workshop at EMNLP 2018

Implicit Negative Feedback in Clinical Information Retrieval
Lorenz Kuhn, Carsten Eickhoff
Medical Information Retrieval Workshop at ACM SIGIR 2016

Experience

I wrote my Master’s Thesis on pruning and generalization in deep neural networks in collaboration with Prof. Yarin Gal at the University of Oxford and Prof. Andreas Krause at ETH Zürich.

Recently, at Cohere, I researched the impact of data set composition and training hyper-parameter choices on the performance of very large language models.

Previously, I obtained a MSc in CS from ETH Zürich, and a BSc in CS from ETHZ and Imperial College London. During my studies, I worked with Prof. Carsten Eickhoff and Prof. Ce Zhang.

I undertook research on medical recommendation systems at IBM Research and ETHZ, and worked as a data scientist for BCG Gamma and QantEv, an Entrepreneur First-backed InsureTech start up.

News

January 2024 — I’ve joined OpenAI full-time.

July 2023 — I’ve joined OpenAI as a resident.

June 2023 — CLAM: Selective Clarification for Ambiguous Questions with Large Language Models was accepted at ICML 2023 Workshop on Deployment Challenges for Generative AI.

May 2023 — Reducing Hallucinations in Generative Language Models via Uncertainty Estimation has been accepted as a poster at the UK AI Fellows Conference 2023.

April 2023 — I gave invited talks on Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation at an ETHZ and TUM seminar, at IST Lisbon and the Intuit AI Seminar.

March 2023 — I served as a reviewer for ICML 2023.

January 2023 — Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation was accepted as a Spotlight paper at ICLR 2023.

November 2022 — Semantic Uncertainty: Linguistic Invariances for Uncertainty Estimation in Natural Language Generation was accepted at the ML Safety Workshop at NeurIPS 2022.