Nils Blank

PhD Researcher · KIT

I’m a PhD researcher at KIT’s Intuitive Robots Lab, advised by Rudolf Lioutikov.

My research focuses on foundation models for robotics, spatial grounding, and learning from robot demonstrations. I study how vision-language models can annotate and curate robot datasets and provide supervision for robot learning.

I received my master’s degree in Computer Science from KIT. During my studies, I interned at SAP SE and IONOS.

Publications

SPARC pipeline: robot demonstrations become spatial annotations with reliability scores, supporting spatial reasoning and robot manipulation.

SPARC: Reliable Spatial Annotations from Robot Demonstrations at Scale

Nils Blank, Paul Mattes, Maximilian Xiling Li, Jakub Suliga, Thomas Roth, Moritz Reuss, Pankhuri Vanjani, Rudolf Lioutikov

CoRL 2026

Turning robot demonstrations into reliable spatial supervision, with explicit control over annotation quality at scale.

Best Paper Award @ Lab2Production Workshop RSS 2026

SIR architecture: observations are converted to sparse scene graphs, graph embeddings, and robot actions.

SIR: Structured Image Representations for Explainable Robot Learning

Paul Mattes, Jan Schwab, Jens Bosch, Maximilian Xiling Li, Nils Blank, Minh-Trung Tang, Moritz Haberland, Rudolf Lioutikov

CVPR 2026

Sparse, task-relevant scene graphs make robot policies interpretable and more robust to visual distractors.

BEAST encodes continuous robot action sequences using compact B-spline tokens.

BEAST: Efficient Tokenization of B-Splines Encoded Action Sequences for Imitation Learning

Hongyi Zhou, Weiran Liao, Xi Huang, Yucheng Tang, Fabian Otto, Xiaogang Jia, Xinkai Jiang, Simon Hilber, Ge Li, Qian Wang, Ömer Erdinç Yağmurlu, Nils Blank, Moritz Reuss, Rudolf Lioutikov

NeurIPS 2025

Compact B-spline action tokens enable fast parallel decoding and smooth robot trajectories, without separate tokenizer training.

NILS combines foundation models to identify objects, understand state changes, and label robot demonstrations in natural language.

Scaling Robot Policy Learning via Zero-Shot Labeling with Foundation Models

Nils Blank, Moritz Reuss, Marcel Rühle, Ömer Erdinç Yağmurlu, Fabian Wenzel, Oier Mees, Rudolf Lioutikov

CoRL 2024

Automatically segmenting and labeling over 115,000 trajectories from 430+ hours of robot demonstrations with pretrained foundation models.