Papers

See my Google Scholar profile for the (likely) most up-to-date list.

  1. Findeis, Arduin. “Evaluation techniques for frontier artificial intelligence.” PhD thesis, 2026. (pdf)
  2. Clifford, Eleanor, Michael Amir, Arduin Findeis, Aaron Zhao, and Robert Mullins. “Open Problems in Constitutional Preference Reconstruction.” arXiv preprint arXiv:2606.30116 (2026). (pdf)
  3. Trhlik, Filip, Aoife O’Flynn, Angela Yu, Arduin Findeis, and Paula Buttery. “LLMs Contain Multitudes: How Deployment Context Reshapes Model-Level Preferences and Values.” arXiv preprint arXiv:2606.13944 (2026). (pdf)
  4. Findeis, Arduin, Timo Kaufmann, Eyke Hüllermeier, and Robert Mullins. “Feedback Forensics: A Toolkit to Measure AI Personality.” preprint arXiv:2509.26305. 2025. (pdf)
  5. Findeis, Arduin, Floris Weers, Guoli Yin, Ke Ye, Ruoming Pang, and Tom Gunter. “Can External Validation Tools Improve Annotation Quality for LLM-as-a-Judge?” In Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (ACL). 2025. (pdf)
  6. Findeis, Arduin, Timo Kaufmann, Eyke Hüllermeier, Samuel Albanie, and Robert D. Mullins. “Inverse Constitutional AI: Compressing Preferences into Principles.” In The Thirteenth International Conference on Learning Representations (ICLR). 2025. (pdf)
  7. Ruis, Laura, Arduin Findeis, Herbie Bradley, Hossein A. Rahmani, Kyoung Whan Choe, Edward Grefenstette, and Tim Rocktäschel. “Do LLMs selectively encode the goal of an agent’s reach?.” In First Workshop on Theory of Mind in Communicating Agents at the Fortieth International Conference on Machine Learning (ICML). 2023. (pdf)
  8. Findeis, Arduin, Fiodar Kazhamiaka, Scott Jeen, and Srinivasan Keshav. “Beobench: a toolkit for unified access to building simulations for reinforcement learning.” In Proceedings of the Thirteenth ACM International Conference on Future Energy Systems (ACM e-Energy), pp. 374-382. 2022. (pdf)

Blog posts

Quite a few over the years, see my blog page.

Python packages
Feedback Forensics (GitHub, app, docs, web) (started 2025)
An open-source toolkit to measure AI personality: track (1) personality changes encouraged by your feedback datasets, and (2) personality traits exhibited by your AI models.
Inverse Constitutional AI (GitHub, paper) (started 2024)
A Python library that implements the Inverse Constitutional AI (ICAI) method, compressing pairwise preference data into a short constitution of principles.
Bauwerk (GitHub, docs) (started 2022)
A meta reinforcement learning (meta RL) benchmark with building control environments. Bauwerk aims to facilitate the development of methods that generalise across buildings to help scale greener building controllers to more buildings.
Beobench (GitHub, docs, paper) (started 2021)
A tool providing easy and unified access to building control environments for reinforcement learning (RL) aiming to enable better comparability and evaluation.
GeoGraph (GitHub, docs) (started 2020)
A tool for analysing habitat fragmentation and related problems in landscape ecology. GeoGraph builds a geospatially referenced graph from land cover or field survey data and enables graph-based landscape ecology analysis as well as interactive visualizations.

Web apps
KraspAI Kompass (no longer active) (started 2023)
A tool to make AI capabilities more human interpretable. Understanding new models’ capabilities is hard. Kompass aims to make it easy by providing tiny but highly informative benchmarks.
langlabel (no longer active) (started 2023)
A tool to automatically label language data using the latest language models. No coding required.