Classifying Interpretive Canons at the Sentence Level: A Benchmark from the German Federal Constitutional Court (September 2026)
Accepted at the ICML 2026 Workshop on AI for Law. Introduces an expert-annotated sentence-level benchmark for classifying interpretive canons in German Federal Constitutional Court decisions and evaluates four LLMs using expert-written and automatically optimized prompts.
Quantitative claims reproduced using the approach outlined in this blog post.
LLMs for Large-Scale Data Collection (July 2026)
Taught the GLEA pre-conference workshop at the Max Planck Institute for the Study of Crime, Security and Law in Freiburg on July 15, 2026. The workshop covered building LLM data collection pipelines, focusing on building evaluations and optimizing against them, with a hands-on case study.
Classifying Interpretive Canons in Federal Constitutional Court Decisions (forthcoming 2026)
To be published in German, titled: Klassifizierung von Auslegungsmethoden in der Rechtsprechung des Bundesverfassungsgerichts, in: Christoph Möllers (ed.), Quantitative Rechtswissenschaft, Mohr Siebeck.
Quantitative Legal Studies (May 2025)
Held a 90-minute lecture on legal argument mining as part of the credit-bearing “Quantitative Legal Studies” elective for J.D.-equivalent and LL.M. students at Freie Universität Berlin and Humboldt University, taught by Professor Andreas Engert and Professor Andreas Fleckner.
Legal Argument Mining: Exploring Interpretive Canon Usage in German Federal Constitutional Court decisions (December 2024)
Research workshop held at Freie Universität Empirical Legal Studies Center (FUELS).
Methods of Leveraging Large Language Models in Empirical Legal Research (June 2024)
Research workshop held at Freie Universität Empirical Legal Studies Center (FUELS).
Determining the minimum base capital present at Annual Shareholders' Meetings (ASMs) (December 2023)
As part of my work at FUELS, I determined the minimum base capital present in Annual Shareholder Meetings (ASMs) of German companies based on unstructured shareholder reports data using an LLM-based classifer.