Bernhard Jaeger

I am a co-founder of KE:SAI, a private non-profit open science AI research lab where I also work. We are hiring.

Research: My research focuses on autonomous driving, which I view as an embodied intelligence problem. My past research has contributed to the development of end-to-end driving technology, which by now is being widely adopted by most self-driving companies. I worked on the TransFuser series of models, which is a widely used baseline in the literature. The latest version, TransFuser v6, is the current (May 2026) state-of-the-art method on all CARLA driving benchmarks. Our survey, "End-to-end autonomous driving: Challenges and frontiers" is the most cited introductory text in the field of end-to-end driving. Recently I have also started using reinforcement learning (RL) for training planning policies. Our method, CaRL, is the first open-source RL policy that outperformed the leading imitation learning methods on the nuPlan benchmark.
Currently, I am interested in closed-loop training, sim2real transfer, and simulation techniques such as Gaussian splatting and causal world models.
I am committed to open contribution to the community. All my papers are freely available on arXiv, and all my code is available on GitHub.

Bio: I studied B.Sc. Informatics: Games Engineering at the Technical University of Munich. Following that, I worked for 1 year as a software developer at FERCHAU as a graphics developer. Afterwards, I did an M.Sc. in Computer Science at the University of Tübingen. From 2022-2026, I worked as a PhD student at the University of Tübingen, as part of the Autonomous Vision Group led by Prof. Andreas Geiger. My research is supported by the International Max Planck Research School for Intelligent Systems.

CV Mail Twitter Bluesky Scholar Github LinkedIn YouTube

Publications

LEAD: Minimizing Learner-Expert Asymmetry in End-to-End Driving
Long Nguyen, Micha Fauth, Bernhard Jaeger, Daniel Dauner, Maximilian Igl, Andreas Geiger, Kashyap Chitta
Conference on Computer Vision and Pattern Recognition (CVPR), 2026
Abs / Paper / Code /

@InProceedings{Nguyen2026CVPR, 
	author = {Long Nguyen and Micha Fauth and Bernhard Jaeger and Daniel Dauner and Maximilian Igl and Andreas Geiger and Kashyap Chitta}, 
	title = {LEAD: Minimizing Learner-Expert Asymmetry in End-to-End Driving}, 
	booktitle = {Conference on Computer Vision and Pattern Recognition (CVPR)}, 
	year = {2026}, 
}

CaRL: Learning Scalable Planning Policies with Simple Rewards
Bernhard Jaeger, Daniel Dauner, Jens Beißwenger, Simon Gerstenecker, Kashyap Chitta, Andreas Geiger
Proc. of the Conf. on Robot Learning (CoRL), 2025
Abs / Paper / Code /

@InProceedings{Jaeger2025CoRL, 
	author = {Bernhard Jaeger and Daniel Dauner and Jens Beißwenger and Simon Gerstenecker and Kashyap Chitta and Andreas Geiger}, 
	title = {CaRL: Learning Scalable Planning Policies with Simple Rewards}, 
	booktitle = {Proc. of the Conf. on Robot Learning (CoRL)}, 
	year = {2025}, 
}

An Invitation to Deep Reinforcement Learning
Bernhard Jaeger, Andreas Geiger
Foundations and Trends® in Optimization, 2024
Abs / Paper /

@article{Jaeger2024FTO, 
	author = {Bernhard Jaeger and Andreas Geiger}, 
	title = {An Invitation to Deep Reinforcement Learning}, 
	year = {2024}, 
	journal = {Foundations and Trends® in Optimization}, 
}

End-to-end Autonomous Driving: Challenges and Frontiers
Li Chen, Penghao Wu, Kashyap Chitta, Bernhard Jaeger, Andreas Geiger, Hongyang Li
Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 2024
Abs / Paper / Code /

@article{Chen2024PAMI, 
	author = {Li Chen and Penghao Wu and Kashyap Chitta and Bernhard Jaeger and Andreas Geiger and Hongyang Li}, 
	title = {End-to-end Autonomous Driving: Challenges and Frontiers}, 
	year = {2024}, 
	journal = {Transactions on Pattern Analysis and Machine Intelligence (T-PAMI)}, 
}

GTA: A Geometry-Aware Attention Mechanism for Multi-View Transformers
Takeru Miyato, Bernhard Jaeger, Max Welling, Andreas Geiger
Proc. of the International Conf. on Learning Representations (ICLR), 2024
Abs / Paper / Code / Website /

@InProceedings{Miyato2024ICLR, 
	author = {Takeru Miyato and Bernhard Jaeger and Max Welling and Andreas Geiger}, 
	title = {GTA: A Geometry-Aware Attention Mechanism for Multi-View Transformers}, 
	booktitle = {Proc. of the International Conf. on Learning Representations (ICLR)}, 
	year = {2024}, 
}

Hidden Biases of End-to-End Driving Models (Winner, 2022 CARLA Challenge Map Track)
Bernhard Jaeger, Kashyap Chitta, Andreas Geiger
Proc. of the IEEE International Conf. on Computer Vision (ICCV), 2023
Abs / Paper / Video / Poster / Code /

@InProceedings{Jaeger2023ICCV, 
	author = {Bernhard Jaeger and Kashyap Chitta and Andreas Geiger}, 
	title = {Hidden Biases of End-to-End Driving Models}, 
	booktitle = {Proc. of the IEEE International Conf. on Computer Vision (ICCV)}, 
	year = {2023}, 
}

TransFuser: Imitation with Transformer-Based Sensor Fusion for Autonomous Driving (Runner Up, 2021 CARLA Challenge)
Kashyap Chitta, Aditya Prakash, Bernhard Jaeger, Zehao Yu, Katrin Renz, Andreas Geiger
Transactions on Pattern Analysis and Machine Intelligence (T-PAMI), 2023
Abs / Paper / Supplemental / Video / Poster / Code /

@article{Chitta2023PAMI, 
	author = {Kashyap Chitta and Aditya Prakash and Bernhard Jaeger and Zehao Yu and Katrin Renz and Andreas Geiger}, 
	title = {TransFuser: Imitation with Transformer-Based Sensor Fusion for Autonomous Driving}, 
	year = {2023}, 
	journal = {Transactions on Pattern Analysis and Machine Intelligence (T-PAMI)}, 
}

Expert Drivers for Autonomous Driving
Bernhard Jaeger
University of Tübingen, 2021
Paper /

@mastersthesis{Jaeger2021Thesis, 
	author = {Bernhard Jaeger}, 
	title = {Expert Drivers for Autonomous Driving}, 
	year = {2021}, 
	school = {University of Tübingen}, 
}

Talks

On output representations for end-to-end driving
Machine Learning for Autonomous Driving Workshop - NeurIPS, 2022

Other activities

One concern that I have as an AI researcher when publishing code is that it can potentially be used in dual-use applications. To solve this, I developed the Civil Software Licenses, which prevent dual-use of open-source software while being minimal in the restrictions they impose.
I am a proponent of rigorous experimental evaluation in autonomous driving research and wrote a guide about common mistakes in the community to help people avoid common pitfalls.
The TransFuser series of models is in version 6 already. I have created a history document for people to have an easier overview.

News

May 01, 2026	I cofounded KE:SAI, a private non-profit open science AI research lab. KE:SAI is a collaboration between Kyutai and the ELLIS Institut Tübingen with the goal of pioneering world-leading research on world models and autonomy.
Oct 26, 2025	I was selected as Top Reviewer at NeurIPS 2025.
Jul 01, 2025	The Vector Stiftung (foundation) supports my research with a grant of 91600 € for the project "Skalierung von verstärkendem Lernen für Ende-zu-Ende Methoden für autonomes Fahren". The grant was competitive, with a 5% acceptance rate (15/300).
Jun 11, 2025	Our team placed 2nd in the Waymo Vision-based End-to-End Driving Challenge held at the CVPR 2025 Workshop on Autonomous Driving.
Jun 11, 2025	Our team placed 3rd in the Scenario Generation Challenge held at the CVPR 2025 Workshop on Autonomous Driving.
Jun 17, 2024	Our team placed 2nd in the CARLA Challenge held at the CVPR 2024 Workshop Foundation Models for Autonomous Systems
Dec 14, 2023	Our team placed 2nd in the CARLA Sensor Track challenge held at the Machine Learning for Autonomous Driving Symposium in New Orleans.
Nov 21, 2022	Our team won the CARLA leaderboard MAP Track challenge at the NeurIPS 2022 Machine Learning for Autonomous Driving workshop.
Apr 01, 2022	I joined the University of Tübingen as a PhD student.
Nov 22, 2021	Our team placed 2nd in the NeurIPS 2021 CARLA Machine Learning for Autonomous Driving autonomous driving challenge .

This website style was created by Michael Niemeyer.
⚛ ⚛ ⚛ ⚛ ⚛ ⚛ ⚛