Tim Franzmeyer

prof_pic.jpg

Hi! I am a fourth-year PhD student at Oxford University, interested in multi-agent systems, reinforcement learning, imitation learning, AI safety and LLMs. I am lucky to be working with Philip Torr, Joao Henriques, and Jakob Foerster.

I’m always interested in meeting new people and in new collaborations. If you’d like to get in touch with me, please email me at frtim at robots dot ox dot ac dot uk.

news

May 07, 2025 New paper out from my Meta internship on preventing LLM Hallucinations with a model-specific finetuning method.
May 01, 2025 Two papers accepted at ICML 2025! One paper with Felipe on attributing LLM answers to either finetuning or pretraining. Second paper is on RL for Quantum Physics, out of a great collaboration with Jan and Aniket.
Nov 04, 2024 Started my internship at Google DeepMind in Zurich. Working on LLM reasoning in the Gemini Post-Training Team with Vikas Yadav, Eric Malmi and Aliaksei Severyn.
Jun 15, 2024 Started my internship at Meta AI in Seattle. Working in the LLama Post-Training Safety Team with Yuning Mao, Luke Zettlemoyer and Madian Khabsa.
Feb 09, 2024 Recent paper on a live LLM benchmark based on Twitter Community notes and Wikipedia Page edits accepted to ACL 2024. Great colab with Suny.
Feb 09, 2024 Linas paper on rethinking out-of-distribution detection in RL will be at AAMAS 2024.
Jan 08, 2024 Two papers accepted at ICLR 2024! One paper on adversarial robustness in RL, the other paper is on imitating desired behaviors from multi-agent observations.
Sep 23, 2023 Joint paper with Felipe on extracting reward functions from Diffusion Models accepted at Neurips 2023!
Sep 20, 2022 Will be presenting my recent paper on Cross-Domain Imitation Learning at NeurIPS 2022.
Jun 14, 2022 Workshop paper on adversarial robustness accepted at ICML 2022.
Jun 10, 2022 New website is live!

selected publications

  1. Measuring the Contribution of Fine-Tuning to Individual Responses of LLMs
    Felipe Pinto Coelho Nuti, Tim Franzmeyer†, and João F Henriques†
    International Conference on Machine Learning (ICML), 2025
  2. Reinforcement Learning for Quantum Control under Physical Constraints
    Jan Ole Ernst*, Aniket Chatterjee*, Tim Franzmeyer*, and Axel Kuhn
    International Conference on Machine Learning (ICML), 2025
    arXiv
  3. HelloFresh: LLM Evaluations on Streams of Real-World Human Editorial Actions across X Community Notes and Wikipedia edits
    Tim Franzmeyer*, Aleksandar Shtedritski*, Samuel Albanie, Philip Torr, João F Henriques, and Jakob N Foerster
    Association for Computational Linguistics (ACL), 2024
    arXiv
  4. Select to Perfect: Imitating desired behavior from large multi-agent data
    Tim Franzmeyer, Edith Elkind, Philip HS Torr, Jakob N Foerster†, and João F Henriques†
    In International Conference on Learning Representations (ICLR), 2024
    arXiv
  5. Illusory Attacks: Information-theoretic detectability matters in adversarial attacks
    Tim Franzmeyer, Stephen McAleer, João F Henriques, Jakob N Foerster, Philip HS Torr, Adel Bibi, and Christian Schroeder Witt
    In International Conference on Learning Representations (ICLR), 2024
    arXiv
  6. Rethinking out-of-distribution detection for reinforcement learning: Advancing methods for evaluation and detection
    Linas Nasvytis, Kai Sandbrink, Jakob Foerster, Tim Franzmeyer†, and Christian Schroeder Witt†
    arXiv preprint arXiv:2404.07099, 2024
  7. Extracting Reward Functions from Diffusion Models
    Felipe Pinto Coelho Nuti*, Tim Franzmeyer*, and João F Henriques
    In Advances in Neural Information Processing Systems (NeurIPS), 2023
    arXiv
  8. Learn what matters: cross-domain imitation learning with task-relevant embeddings
    Tim Franzmeyer, Philip HS Torr, and João F Henriques
    In Advances in Neural Information Processing Systems (NeurIPS), 2022
    arXiv
  9. Learning Altruistic Behaviours in Reinforcement Learning without External Rewards
    Tim Franzmeyer, Mateusz Malinowski, and João F Henriques
    International Conference on Learning Representations (ICLR), 2022
    arXiv
  10. Scalable Biologically-Aware Skeleton Generation for Connectomic Volumes
    Brian Matejek*, Tim Franzmeyer*, Donglai Wei, Xueying Wang, Jinglin Zhao, Kálmán Palágyi, Jeff W Lichtman, and Hanspeter Pfister
    IEEE Transactions on Medical Imaging, 2022
    Code Website