Benjamin Heltzel

I am a Master's student in Robotics, Cognition, and Intelligence at TUM. My current research focuses on physically-grounded world models, 4D computer vision, and robot learning.

Previously, I completed research projects with Prof. Berthold Bäuml, Prof. Angela Dai, and Prof. Matthias Niessner. I received my Bachelor's degree in Engineering Science, specializing in machine learning, from TUM.

In 2018, I co-founded Seismos Network, where we built an earthquake early warning network on crowdsourced phone accelerometer data via neural networks.

Benjamin Heltzel

Research

Towards Sim2Real Transfer in Robotic Reinforcement Learning via Proprioceptive Self-Adaptation
Towards Sim2Real Transfer in Robotic Reinforcement Learning via Proprioceptive Self-Adaptation
Benjamin Heltzel·Leon Sievers·Prof. Berthold Bäuml

We demonstrate real-time proprioceptive adaptation to unknown intrinsic robot dynamics in dexterous in-hand manipulation, paving the way for long-term deployment without re-calibration downtime.

Bachelor Thesis, 2026
AIDX Lab
SAC<sup>2</sup>: Rapid Adaptation to Random Disturbances in Partial Observability via Off-Policy Meta-Reinforcement Learning
SAC2: Rapid Adaptation to Random Disturbances in Partial Observability via Off-Policy Meta-Reinforcement Learning
Benjamin Heltzel·Leon Sievers·Prof. Berthold Bäuml

An asymmetric LSTM-based meta-RL variant of Soft Actor-Critic enables real-time adaptation by inferring hidden task dynamics from history in POMDPs.

Research Project, 2024
AIDX Lab
PromptScene: Adaptive Prompt Learning for Open-Vocabulary 3D Instance Segmentation
PromptScene: Adaptive Prompt Learning for Open-Vocabulary 3D Instance Segmentation
Benjamin Heltzel, Simon Blessmann, Ayaka Nanri
Mohamed El Amine Boudjoghra·Prof. Angela Dai

Adaptive prompt learning in OpenScene with Mask3D class-agnostic instance grouping enables flexible 3D object recognition without predefined labels.

Research Project, 2024
3D AI Lab
Dif-fused DINO-tracker: Zero-shot Point Tracking with Video Diffusion Motion Features
Dif-fused DINO-tracker: Zero-shot Point Tracking with Video Diffusion Motion Features
Benjamin Heltzel·Dr. Lei Li·Prof. Matthias Niessner

Cross-attention fusion of temporal representations from video diffusion models with DINO features improves zero-shot point tracking coherence.

Research Project, 2024
Visual Computing & AI Lab

Startups

Seismos Network
Seismos Network
Co-founder

We built Seismos to provide global earthquake early warning alerts with ML-based real-time earthquake detection on crowdsourced phone accelerometer data. In the US, the app was also slated to deliver USGS ShakeAlert notifications for CA, OR, and WA.

2018-2021
Vietnam, USA

Talks

Neural Networks for Real-time Earthquake Early Warning

Prof. Christian Grosse - TUM Non-destructive Testing Lab,

Prof. Steven Glaser - UC Berkeley Seismo Lab

2024
Doctoral Seminar, TUM

Teaching

Introduction to Deep Learning
Teaching Assistant·Prof. Daniel Cremers
Winter 2024
TUM

Design based on Haoran Geng's website, with source code from Jon Barron and Jekyll fork by Leonid Keselman.
© 2026 Benjamin Heltzel