Alex Laitenberger

Truthfulness, Interpretability, and Language Model Systems

I am a PhD Fellow at the University of Copenhagen, working in the CopeNLU group under the supervision of Pepa Atanasova and Isabelle Augenstein.
My research interests center on understanding how language models represent and use factual information, and how these insights can contribute to more reliable and trustworthy systems. I am particularly interested in truthfulness, interpretability, and the design and understanding of language model systems.
My previous work, conducted in collaboration with Christopher D. Manning and Nelson F. Liu, was published at EMNLP 2025 and studied retrieval-augmented generation for long-context question answering. It showed that strong, structure-preserving RAG baselines can match or outperform more complex multi-stage approaches, and introduced a simple baseline for more rigorous RAG methodology research.

[LinkedIn] [Google Scholar] [GitHub]

---

Publications

Stronger Baselines for Retrieval-Augmented Generation with Long-Context Language Models
Alex Laitenberger, Christopher D. Manning, Nelson F. Liu.
EMNLP 2025 (Main Conference), Suzhou, China, 2025.
[Code] [Poster]

Expanding Horizons in RAG: Exploring and Extending the Limits of RAPTOR
Alex Laitenberger.
Research Project (Stanford CS224N), 2024.
[Poster]

Concept for Discovering Relevant Information through the Automated Analysis of News Articles
Alex Laitenberger.
Bachelor’s Thesis, 2011.

Evaluating Product-Market Fit for a Consumer-Oriented App Startup
Alex Laitenberger.
Master’s Thesis, 2018