Soofi S marks the beginning of the project’s first release phase. The model is designed for organizations that need to run AI applications transparently, adaptively, and on their own or sovereign infrastructure — for example, in industrial processes, the analysis of extensive technical and regulatory documents, code generation, or agentic AI systems. Trained from scratch on 27 trillion (27T) tokens, Soofi S is a 30 billion parameter (30B-A3B) mixture-of-experts model whose hybrid Mamba-Transformer architecture combines high throughput with low energy consumption. Soofi S has been trained primarily on English and German text and achieves top-tier results among open models in its size class in English; in German, it leads the peer group. Soofi S is initially released as a base model, which can already be fine-tuned for specific domains; post-trained variants for dialogue and agentic applications will follow.

“Whoever controls the foundation models controls a central part of future digital value creation and strengthens their sovereignty and resilience. With Soofi, we are building an open foundation on which businesses, SMEs, and the public sector can develop transparent AI applications based on their own data, without becoming permanently dependent on individual non-European models,” says Jörg Bienert, Managing Director of the Center for Sovereign AI, German AI Association.

A particular focus lies on transparency: the consortium will not only release model weights but also publish technical documentation on training methodology, data preparation, and the data pipelines used. This makes Soofi S more auditable for businesses, public authorities, and researchers, and more readily adaptable to specific use cases.

Soofi S and subsequent models are trained on Deutsche Telekom’s Industrial AI Cloud in Munich, using NVIDIA’s open-source AI framework. Initial results show that the Soofi S base model matches or outperforms international models of comparable size across German and English benchmarks.

“Soofi S is not intended as yet another general-purpose chatbot, but as a technical foundation for industrial AI. What matters is that Soofi S performs not only well in benchmarks, but can be deployed reliably, efficiently, and transparently in production,” says Nicolas Flores-Herr, Technical Project Lead for Soofi and Team Lead at Fraunhofer IAIS.

“Digital sovereignty arises where cutting-edge scientific research and industrial practice intersect directly. With ‘Soofi S,’ we are demonstrating that Europe holds the keys to the next generation of AI technologies. The DFKI is contributing its expertise here to develop models that combine performance with transparency. This is a crucial lever for providing the European economy with an independent and future-proof AI infrastructure.”

Prof. Dr. Antonio Krüger, CEO DFKI

At PLDI this year, only 10 papers were given this award out of 115 accepted papers.

DFKI and Inria have been collaborating in a bilateral partnership since 2020. The focus is on joint research projects carried out by mixed teams from both organizations. Two current projects exemplify societal challenges in Germany and France and make a concrete contribution to greater social participation. The “RoGSiLT” and “NEARBY” projects will be presented at the joint booth in the German Park.

RoGSiLT (Robust and Generalizable Sign Language Translation) is developing AI-based solutions for German and French Sign Language (DGS and LSF). Innovative AI methods aim to significantly improve translations between spoken language and sign language. The goal is to improve both the translation of text into sign language and the conversion of sign language from videos into written language. Modern techniques such as multimodal neural networks, self-supervised learning, and large language models are intended to overcome existing hurdles—such as limited data availability, lack of generalizability, and unnatural translations. A key component is the development of new data resources, including extensive parallel corpora of sign language videos and associated texts.

Project presentation: Wednesday, June 17, all day, Hall 7.3, Booth 3E14

Brain-computer interfaces (BCIs) open new avenues for human-machine interaction by using brain signals to directly control technical systems. NEARBY (Noise and variability-free BCI systems for out-of-the-lab use) develops innovative BCI systems with low noise and variability that function reliably even outside the laboratory. The goal is to create robust, practical solutions that pave the way for the use of brain-computer interfaces in everyday life—for greater self-determination, efficiency, and intuitive interaction. At the conclusion of the project, the German-French team will present the current state of research.

Project presentation: Thursday and Friday, June 18–19, all day, Hall 7.3, Booth 3E14
 

DFKI and Inria are taking their partnership to a new level: The leading research institutions will sign an agreement to establish an open, binational German-French research center for artificial intelligence.

The signing ceremony will be held in the presence of Dorothee Bär, Federal Minister for Research, Technology, and Space, and Philippe Baptiste, French Minister for Higher Education, Research, and Space. This ambitious project aims to establish a strong European AI player at the intersection of cutting-edge research, industry, and society.

German Park Stage, June 18, 10:00–10:30 a.m.

www.vivatech.com
https://www.dfki.de/en/web/qualifications-networks/international-cooperation/german-french

https://french-german-ai-center.org
 

Graduates with a Bachelor’s degree in physics, mathematics or computer science can apply for a place on the programme until 30 June.

Quantum computing is widely seen as one of the key technologies of the 21st century. The global development of powerful quantum hardware systems continues to gather pace and with it grows the demand for new algorithms, software concepts and cryptographic methods. And this is where the new Master’s programme comes in. With its focus firmly on the software side of quantum computing, it deals with the theoretical foundations of information processing in quantum systems. ‘Quantum information theory is, in a sense, the theoretical underpinning of the software used in quantum computers. It is rooted equally within the fields of mathematics, computer science and physics, and calls for an integrated and interdisciplinary perspective,’ explains Professor Moritz Weber, Scientific Director at the Center for Quantum Technologies (QuTe) at Saarland University, who has been instrumental in designing the new degree programme.

Compared with other study programmes in the quantum technologies sector, many of which are strongly physics-focused, the Saarbrücken Master’s programme is based at the Department of Mathematics and focuses particularly on mathematical and computer science aspects. The M.Sc. programme in Saarbrücken therefore offers a route into the world of quantum research even for students with no prior background in physics. ‘At its core, the programme addresses the fundamental questions raised by the development of quantum computing technology: How do you actually program a quantum computer? In what way does it function differently from a classical computer? What characterizes quantum computing and what does that have to do with mathematics,’ says Moritz Weber. Students on the programme acquire a solid grounding in areas such as quantum algorithms and the mathematical foundations of quantum information. Mandatory elective modules allow students to specialize in areas such as quantum complexity, quantum error correction or other research fields of current interest.

A particularly attractive feature of the Master’s programme is that students gain early exposure to the questions of current research relevance. This is possible because of the close links that exist between the study programme and the recently established Center for Quantum Technologies, and the collaborative ties with other leading research institutes in Germany, including the Helmholtz Center ‘Forschungszentrum Jülich’. Students can also gain practical experience through a supervised internship – either in a research or industry setting. The programme also offers flexible study pathways, enabling students to combine quantum computing equally with either classical computer science or mathematics. This flexibility allows them to tailor their studies to their individual interests and longer-term career aims.

Graduates from the programme will be highly sought-after specialists in a very innovative and fast-growing sector. Career opportunities range from research and development to the quantum computing industry and the IT and cybersecurity sectors, for example in post-quantum cryptography. Further prospects include professional roles in software development, data analysis and high-performance computing. At the same time, the programme provides targeted preparation for an academic career, including doctoral study in the broader field of quantum science.

The Master’s programme is taught entirely in English and has been intentionally designed to appeal to international students as well. The standard period of study is four semesters, and the programme leads to a Master of Science degree. The deadline for applications for the coming winter semester is 30 June.

Further information and details about how to apply on this Website.

Thanks to €48 million in support from European and national funding sources, the RICAIP (Research and Innovation Centre on Advanced Industrial Production) project has created an interconnected ecosystem of industrial testbeds at CIIRC CTU and CEITEC BUT in Brno, linked with leading German research centres DFKI and ZeMA, as well as additional partner institutions across Europe. The testbeds enable systematic development and testing of new approaches in industrial AI, robotics, distributed manufacturing, and advanced automation under conditions close to real-world industrial operations, significantly strengthening the ability to transfer research results into industrial practice.

“RICAIP is a truly European effort. At the intersection of education, research, innovation and industrial development, we have built a technological infrastructure and a network of teams in the Czech Republic and German that can be a leader in the future of Industry 4.0. We have set up the RICAIP testbeds and filled them with an ecosystem of collaboration with many stakeholders. At the same time, it is an ongoing process of setting up a sustainable format as the RICAIP Centre and continuing to expand European cooperation.”

Dr. Tilman Becker, RICAIP Director

The conference opened a broader discussion on how intelligent physical systems, autonomous robotics, and AI-driven manufacturing will shape the future of European industry. Keynote speakers included Prof. Wolfgang Wahlster from DFKI, one of the founding figures of Industry 4.0, and Prof. Duncan McFarlane from the University of Cambridge, a leading expert in industrial intelligence systems and digital twins. The programme also featured Valentina Ivanova, Deputy Director for European and International Affairs at CEA-List and coordinator of the European AI-MATTERS initiative focused on testing and experimentation facilities for AI in manufacturing. Their contributions provided a forward-looking perspective on industrial AI, robotics, and intelligent physical systems in a European context.

The interaction between artificial intelligence and the physical world is one of the key scientific and technological challenges of our time, with a profound impact on the future of manufacturing, logistics, and other industrial sectors,” said Prof. Vladimír Mařík, Scientific Director of CIIRC CTU and principal investigator of the RICAIP project. “RICAIP has created a strong environment for the development of the Czech AI ecosystem and has significantly contributed to new initiatives such as the Czech AI Factory“.

According to Prof. Mařík, close cooperation between Czech institutions and German partners DFKI and ZeMA has played a crucial role. “Thanks to this collaboration, CIIRC CTU in Prague and CEITEC BUT in Brno are now recognised as respected European centres of excellence in industrial AI, also involving VSB – Technical University of Ostrava. At the same time, we are working intensively to ensure the long-term sustainability of both the infrastructure and research potential, as European industry will require strong scientific and technological support during the ongoing digital transformation.” 

“Collaboration initiated within RICAIP contributes to European digital sovereignty in the field of industrial artificial intelligence. Close cooperation and the shared use of state-of-the-art test environments have created a European innovation ecosystem that enables us to independently develop, scale, and deploy key technologies in industrial AI.”

Prof. Antonio Krüger, CEO DFKI

The importance of interconnected testbeds was also highlighted by representatives of individual sites. “The future of industrial AI is not created in isolated laboratories, but in interconnected research infrastructures where technologies, expertise, and experiments can be shared across countries,” said Khansa Rekik from ZeMA. “Within RICAIP, we were able to connect locally developed robotic and AI solutions into broader distributed manufacturing scenarios.”

“RICAIP represented a major qualitative leap for our testbed,” added Dr. Pavel Burget, Director of the RICAIP Testbed Prague. “We have created an environment where cutting-edge research meets real industrial applications – from autonomous robotic manipulation and digital twins to AI-based quality inspection in extremely short production cycles.” 

“Long-term international cooperation and shared research infrastructures are key to developing future technologies. For our institute, participation in such initiatives is an opportunity not only to advance research, but also to transfer its results into industrial practice and strengthen European competitiveness in AI and advanced manufacturing,” confirmed Prof. Radimír Vrba, Director of CEITEC BUT in Brno.

The importance of cooperation between research and industry was also emphasised by representatives of the industrial sector. “Strengthening Europe’s competitiveness requires much closer cooperation between research and industry. Initiatives such as collaboration between companies, CIIRC CTU, and infrastructures like RICAIP clearly demonstrate how research results can be transformed into real industrial value. Digitalization, automation, and artificial intelligence will be the key drivers of transformation in European industry in the coming decade. At the same time, we need an environment that motivates companies to invest in research and innovation,” said Martin Jahn, Member of the Board for Sales and Marketing at Škoda Auto, Vice-President of the Confederation of Industry of the Czech Republic, and President of AutoSAP.

“If Europe is to remain competitive on the global stage, we cannot innovate in isolation,” added Eduard Palíšek, CEO of Siemens Czech Republic. “Close collaboration between industry and academia, such as that represented by CIIRC CTU and the RICAIP testbeds, allows us not only to validate new technologies in real industrial environments, but also to jointly address challenges such as manufacturing resilience and cybersecurity. True competitiveness is not built on new technologies alone, but also on the courage to share know-how and push the boundaries of what digitalization makes possible.“ 

RICAIP Days 2026 thus represented not only the conclusion of a successful European project, but above all the beginning of a new phase of European cooperation in industrial AI, intelligent physical systems, and technology transfer between research and industry.

The conference was followed by Tech Dating 2026, an open day at CIIRC CTU for companies, organised in cooperation with EDIH CTU, AI-MATTERS, the National Centre for Industry 4.0, CzechInvest, and other partners. It offered hands-on consultations, technology demonstrations, and the opportunity to discuss concrete challenges in AI, automation, and digitalisation directly with research teams from CIIRC CTU and partner institutions.

The Fonds der Chemischen Industrie (FCI) is funding the project led by Professors Tanja Gulder and Andrea Volkamer to the tune of €80,000 as part of a special funding programme designed to make data science a permanent part of chemistry degree programmes. 

The following text has been machine translated from the German with no human editing.

Whether substances are being analysed, experiments conducted, new active compounds tested or theoretical models examined: everything that takes place in a chemistry laboratory generates data. Without these measurements and observations, nothing works – if they are not recorded, the results of even the best work are lost. What has been true since the dawn of Chemistry is even more so today. “Data and its processing are an indispensable foundation for research and development. New computer-aided methods, including artificial intelligence, are creating new ways to achieve research results using large amounts of data and to make them usable for industry and society,” explains Tanja Gulder, Professor of Organic Chemistry at Saarland University. She researches and develops sustainable chemical processes modelled on nature, including for novel and improved active substances, using digital computer methods. 
“Technology is significantly transforming the chemistry and pharmaceutical industries. This is also changing the demands placed on the specialists working in these fields. Industry is looking for chemists who possess knowledge of data, its machine- -based processing and the use of AI,” says Tanja Gulder. Chemistry education must catch up and integrate digital methods into teaching from the very start. “So far, the Chemistry curriculum in Germany has not covered this area sufficiently. Even some PhD students today are reluctant to work with data. That is why we want to introduce digital content into Chemistry degree programmes using a new teaching concept, starting with the Bachelor’s degree and continuing throughout to the Master’s degree,” explains Gulder. 

To this end, the chemist at Saarland University is collaborating with Andrea Volkamer, Professor of Data-Driven Drug Design. The computational chemist develops computer-based methods—encompassing both algorithmic approaches and AI models—to predict which drug candidates are the most promising; she is also a specialist in digital teaching methods. “Knowledge of the use of artificial intelligence needs to be better integrated into the curriculum in general,” emphasises Andrea Volkamer.

The two researchers are now receiving funding for their project under the special funding programme of the Chemical Industry Fund (FCI): the Saarbrücken application was successful in a nationwide competition alongside 22 other universities and colleges. Numerous institutions had submitted applications for funding. The Saarbrücken project concept will receive 80,000 euros in funding over three years. In total, the fund is investing 1.6 million euros in data science within Chemistry degree programmes to embed innovative teaching concepts on AI, big data and laboratory automation in higher education.

Over the next three years, Gulder and Volkamer will develop a teaching concept with their teams. “The concept is intended to serve as a sustainable model,” says Andrea Volkamer. Its modules will initially be developed within the Department of Organic Chemistry at Saarland University, but they are intended to be applicable beyond the boundaries of this discipline – generally across the life sciences, such as in Pharmaceutical Science or biotechnology – and beyond the Saarbrücken campus to other universities. 

The new concept for the introductory bachelor’s practical course aims to teach students the basics of working with data and data management – in other words, how to generate data as the foundation for machine learning and how to store it in a way that is both useful and retrievable by third parties. “We also want to teach students the ‘FAIR’ principles of research data management,” explains Tanja Gulder. FAIR stands for the initial letters of the English terms ‘findable’, ‘accessible’, ‘interoperable’ and ‘reusable’. “The advanced practical course will, among other things, focus on introducing students to working with databases such as electronic lab notebooks and providing them with a basic understanding of AI, machine learning and programming languages. The aim is to equip them with the tools to analyse large volumes of data, identify complex patterns and correlations, optimise synthesis routes or predict reaction outcomes,” explains the chemist.

Gulder, who is involved in two major collaborative research centres and two Research Training Groups run by the German Research Foundation (DFG), is working alongside colleagues at the University of Leipzig to set up another major research project in the field of digital Chemistry: the aim here will be to advance chemical research for drug discovery using modern computer-aided methods, including artificial intelligence. “With this new teaching concept, we are also training the next generation of chemists who will be working on this research project,” says Gulder, explaining her long-term goals.

The Chemical Industry Fund was established in 1950 and is the funding body of the German Chemical Industry Association. The Faculty of Natural Sciences and Technology and the Chemistry department at Saarland University are contributing an additional 20 per cent of the funding as an investment in the quality of teaching.

For further information:

Prof. Dr Tanja Gulder: Email: tanja.gulder@uni-saarland.de

Prof. Dr Andrea Volkamer: Email: volkamer@cs.uni-saarland.de

Press photos available for download:
Press photos can be found on this news website: 

https://www.uni-saarland.de/aktuell/ki-im-Chemistry-studium-47017.html
You may use the press photos free of charge in connection with this press release and reporting on Saarland University, provided you credit the photographer.

‘Software solutions are embedded in our everyday lives, and when they are faulty the consequences can be serious. So, it‘s essential that programmers understand their code and don‘t overlook errors or introduce new ones when adding further functionality,’ says Sven Apel, professor of computer science at Saarland University. Apel and his colleagues want to understand more precisely what happens in a software developer’s brain when writing and analysing code. Three years ago, he brought Axel Mecklinger, professor of experimental neuropsychology at Saarland University, into the project. Together, they combined electroencephalography (EEG) with eye-tracking data to record signals known technically as fixation-related potentials (FRPs). ‘The advantage of the FRP approach is that it allows us to record brain activity at the exact moment when the eyes stop moving and focus on a specific target,’ explains Axel Mecklinger.

The research team set out to discover how software developers react when they encounter confusing snippets of code known as ‘atoms of confusion’. These small portions of code occur quite frequently in source code and cause a person and a machine to come to different conclusions regarding the output. While the computer can interpret and execute them unambiguously, they are not intuitively clear to the programmer, which means the programmer may misunderstand how the program works. Anna-Maria Maurer, a doctoral research student in computer science working with Professor Apel, incorporated this type of confusing code into the experimental design. She recruited 24 programmers as participants, whose brain activity and eye movements were recorded over around 1,700 trials.

To analyse the measurement data, the team drew on methods and expertise from psycholinguistics, although the methodology could not simply be transferred to the study of software programming. Earlier studies had already shown that programming activates brain regions similar to those involved in natural language processing, but the way programmers approach code differs from the way people process language. ‘When we want to understand how the brain processes particular conversational situations, we ask participants to read short text passages and compare this with the EEG and eye-tracking data. But when reading code, programmers process larger contextual blocks, scanning several lines at once and perceiving complex structures as single units,’ explains Vera Demberg, professor of computational linguistics, who with her team was involved in analysing the data. To take this added complexity into account, the experimental set-up and design had to be correspondingly more sophisticated. The code snippets were presented to the participants in three thematic blocks, with each block comprising 24 individual trials, while EEG and eye-movement data were recorded and synchronized to the millisecond.

When the team compared EEG signals from earlier natural language studies with the new findings from their software programming study, they identified a striking pattern known in neuropsychology as late frontal positivity. ‘When the programmers encountered confusing snippets of code, they showed brain activity similar to that seen in linguistic experiments where participants read sentences containing unexpected turns of phrase. The brain then adapts in a split second, checking the information against long-term memory and updating the mental representation of the new situation in order to make sense of it,’ explains Vera Demberg. To illustrate what’s happening, Axel Mecklinger cites the sentence ‘Theo wants to chop wood, so he goes to fetch a jacket.’ as an example of just such an unexpected turn in conversation. Upon encountering the words ‘goes to fetch’ the reader would normally expect this to be followed by ‘an axe’. A jacket, by contrast, is certainly plausible, but still comes as a surprise in this context. ‘In our EEG experiments on language processing, unexpected words such as “jacket” generate a late frontal positivity – a signal that bears a very strong resemblance to the EEG response elicited by confusing code snippets,’ says neuropsychologist Axel Mecklinger.

‘Because programmers spend 70 to 80 percent of their time trying to understand code, it’s important that we understand how their thought processes work. The insights we gain can help us develop better tools that either eliminate coding pitfalls from the outset or make them easier to detect. These findings could also inform how we go about training software developers,’ explains computer scientist Sven Apel. In future studies, he hopes to investigate whether programmers show different patterns of brain activity when the confusing snippets of code are actually faulty, or when they are shown lines of code that do not require spontaneous rethinking, i.e. spontaneous revision of the programmer’s mental representation of the situation.

The study, published in the prestigious journal Scientific Reports, involved Annabelle Bergum, Anna-Maria Maurer, Norman Peitek, Regine Bader, Axel Mecklinger, Janet Siegmund, Vera Demberg and Sven Apel. All of the authors are researchers at Saarland University, with the exception of Janet Siegmund, who is professor of software engineering at Chemnitz University of Technology. The study is linked to several major research programmes at Saarland University and received funding from them. These include the Transregional Collaborative Research Centre 248, ‘Foundations of Perspicuous Software Systems’ (co-spokesperson: Professor Holger Hermanns), the ERC Advanced Grant ‘Brains on Code’ (PI: Professor Sven Apel), and the Collaborative Research Centre 1102 on Information Density and Linguistic Encoding (spokesperson: Professor Elke Teich), in which Regine Bader, Vera Demberg and Axel Mecklinger are involved.

Original publication:
Annabelle Bergum, Anna-Maria Maurer, Norman Peitek, Regine Bader, Axel Mecklinger, Vera Demberg, Janet Siegmund and Sven Apel, Fixation-related potentials reveal that confusing program code elicits a late frontal positivity. In: Scientific Reports 16, 16833 (2026): https://doi.org/10.1038/s41598-026-50946-9 

Further information:
Chair of Software Engineering: https://www.se.cs.uni-saarland.de 

Questions can be addressed to:
Professor Sven Apel
Chair of Software Engineering
Saarland University
Tel.: +49 681 302-57211
Email: apel(at)cs.uni-saarland.de 

The IEEE/CVF Conference on Computer Vision and Pattern Recognition, or CVPR for short, is one of the most important conferences in the field of computer vision research and took place this year from 3 to 7 June in Denver. The DFKI was represented there with several accepted papers from various areas of research. The focus was on a paper from the Augmented Vision research group that addresses a key weakness in current 3D scene analysis: systems recognise objects but often fail to understand how they relate to one another.

The DFKI contributions to the main conference cover a broad spectrum of visual AI. From the Augmented Vision research area come “ReLaGS: Relational Language Gaussian Splatting”, “DriverGaze360: OmniDirectional Driver Attention with Object-Level Guidance”, “LiREC-Net: A Target-Free and Learning-Based Network for LiDAR, RGB, and Event Calibration”, and “SIMSPINE: A Biomechanics-Aware Simulation Framework for 3D Spine Motion Annotation and Benchmarking”. 

These are joined by “OpenMarcie: A dataset for multimodal action recognition in industrial environments” and “When Pretty Isn’t Useful: An investigation into why modern text-to-image models fail as reliable generators of training data”, as well as “YieldSAT: A multimodal benchmark dataset for high-resolution yield prediction” from Kaiserslautern, “Synthesising Visual Concepts as Vision-Language Programs” from Darmstadt, and “SceMoS: Scene-Aware 3D Human Motion Synthesis by Planning with Geometry-Grounded Tokens” from Saarbrücken. Collectively, the topics range from open 3D scene understanding and multimodal perception to sensor calibration, medical simulation, synthetic training data and generative motion modelling.

Within this spectrum, ReLaGS stands out. The paper by Yaxu Xie, Abdalla Arafa, Alireza Javanmardi, Christen Millerdurai, Jia Cheng Hu, Shaoxiang Wang, Alain Pagani and Didier Stricker combines a hierarchical 3D scene representation with an explicit scene graph that models relationships between objects. This makes it possible not only to identify objects within a scene, but also to process relational queries such as ‘the cup next to the laptop’ or more nuanced part-whole relationships within complex 3D environments.

The method is based on Gaussian splatting, a state-of-the-art technique for high-resolution 3D reconstruction. ReLaGS supplements this with linguistic semantics and relational reasoning, organises scenes hierarchically – from parts through objects to the entire space – and does not require scene-specific training.

“With ReLaGS, we have shown that 3D scene understanding need not stop at the recognition of individual objects. The key lies in modelling relationships, hierarchies and semantic contexts together – only then does reconstruction truly become machine understanding.”

In the paper, the researchers report that ReLaGS generates a complete scene graph in under 15 minutes and renders at over 200 frames per second. Compared to RelationField, the approach is therefore 4.7 times faster and 7.6 times more memory-efficient. On benchmarks for open 3D segmentation, scene graph prediction and relation-guided instance segmentation, ReLaGS also achieves state-of-the-art results. 

This is relevant to research because 3D scene understanding is increasingly needed in areas where machines are required to operate safely and contextually in complex environments: in robotics, XR, industrial digital twins, or semantically rich human-machine interfaces. ReLaGS demonstrates how geometric reconstruction, linguistic semantics and relational structure can be integrated into a single framework.

In addition to the Main Conference, DFKI was also represented in other formats at CVPR 2026. From the Augmented Vision research group, “GHOST: Fast Category-Agnostic Hand-Object Interaction Reconstruction from RGB Videos Using Gaussian Splatting” and “ReConText3D: Replay-based Continual Text-to-3D Generation” were accepted as findings posters. “TAUE: Training-free Noise Transplant and Cultivation Diffusion Model” was also featured among the Findings posters. 

In addition, there were workshop presentations entitled ‘Probing the Reliability of Driving VLMs: From Inconsistent Responses to Grounded Temporal Reasoning’ at the AUTOPILOT workshop, and ‘Inpaint360GS: Efficient Object-Aware 3D Inpainting via Gaussian Splatting for 360° Scenes’ at the SPAR-3D workshop. This demonstrated DFKI’s presence at CVPR 2026 not only at the main conference but also in formats where current methodological issues and new fields of application are discussed.

• ReLaGS: Relational Language Gaussian Splatting – Yaxu Xie, Abdalla Arafa, Alireza Javanmardi, Christen Millerdurai, Jia Cheng Hu, Shaoxiang Wang, Alain Pagani, Didier Stricker 
• DriverGaze360: OmniDirectional Driver Attention with Object-Level Guidance – Shreedhar Govil, Didier Stricker, Jason Rambach 
• LiREC-Net: A Target-Free and Learning-Based Network for LiDAR, RGB, and Event Calibration – Aditya Ranjan Dash, Ramy Battrawy, René Schuster, Didier Stricker 
• SIMSPINE: A Biomechanics-Aware Simulation Framework for 3D Spine Motion Annotation and Benchmarking – Muhammad Saif Ullah Khan, Didier Stricker 
• OpenMarcie: Dataset for Multimodal Action Recognition in Industrial Environments – Hymalai Bello, Lala Ray, Joanna Sorysz, Sungho Suh, Paul Lukowicz 
• When Pretty Isn’t Useful: Investigating Why Modern Text-to-Image Models Fail as Reliable Training Data Generators – Krzysztof Adamkiewicz, Brian Moser, Stanislav Frolov, Tobias Christian Nauen, Federico Raue, Andreas Dengel 
• SceMoS: Scene-Aware 3D Human Motion Synthesis by Planning with Geometry-Grounded Tokens – Anindita Ghosh, Vladislav Golyanik, Taku Komura, Philipp Slusallek, Christian Theobalt, Rishabh Dabral 
• GHOST: Fast Category-Agnostic Hand-Object Interaction Reconstruction from RGB Videos Using Gaussian Splatting – Ahmed Tawfik Aboukhadra, Marcel Rogge, Nadia Robertini, Abdalla Arafa, Jameel Malik, Ahmed Elhayek, Didier Stricker 
• Probing the Reliability of Driving VLMs: From Inconsistent Responses to Grounded Temporal Reasoning – Chun-Peng Chang, Chen-Yu Wang, Holger Caesar, Alain Pagani 
• Inpaint360GS: Efficient Object-Aware 3D Inpainting via Gaussian Splatting for 360° Scenes – Shaoxiang Wang, Shihong Zhang, Christen Millerdurai, Rüdiger Westermann, Didier Stricker, Alain Pagani 
• ReConText3D: Replay-based Continual Text-to-3D Generation – Muhammad Ahmed Ullah Khan, Muhammad Haris Bin Amir, Didier Stricker, Muhammad Zeshan Afzal
• TAUE: Training-free Noise Transplant and Cultivation Diffusion Model – Daichi Nagai, Ryugo Morita, Shunsuke Kitada, Hitoshi Iyatomi
• YieldSAT: A Multimodal Benchmark Dataset for HighResolution Crop Yield Prediction – Miro Miranda, Deepak Pathak, Patrick Helber, Benjamin Bischke, Hiba Najjar, Francisco Mena, Cristhian Sanchez, Akshay Pai, Diego Arenas, Matias Valdenegro-Toro, Marcela Charfuelan, Marlon Nuske, Andreas Dengel
• Synthesizing Visual Concepts as Vision-Language Programs – Antonia Wüst, Wolfgang Stammer, Hikaru Shindo, Lukas Helff, Devendra Singh Dhami, Kristian Kersting

Jointly presented by the ACM Symposium on Principles of Distributed Computing (PODC) and the EATCS Symposium on Distributed Computing (DISC), the prize recognizes work that has had significant impact on the theory and practice of distributed computing over at least a decade. This years awardees are Atish Das Sarma, Stephan Holzer, Liah Kor, Amos Korman, Danupon Nanongkai, Gopal Pandurangan, David Peleg, and Roger Wattenhofer. Their paper shows that in distributed graph computations, the data links between machines form the key bottleneck. The system’s performance is thus fundamentally constrained by the bandwidth of these connections. This insight helped to define the modern theory of bandwidth-constrained distributed computing, as mentioned in the award citation.

Originally published at the ACM Symposium on Theory of Computing (STOC) in 2011 and later in the SIAM Journal on Computing in 2012, the paper introduced a framework to determine the fundamental limits of distributed computing, a method where multiple computing systems act as a single unit. By connecting communication theory with network verification, the authors created a systematic way to prove the minimum time required to solve various problems. This applied to many key tasks, such as finding the shortest path or minimum cuts. Today, these methods are standard tools in the field. Also, the insights presented in the paper have shaped subsequent research, such as extending the framework to new problems and designing algorithms with matching bounds.

The 2026 Dijkstra Prize Committee comprises James Aspnes (Yale University), Keren Censor-Hillel (Technion), Cyril Gavoille (University of Bordeaux), Seth Gilbert (National University of Singapore), Andrzej Pelc (Université du Québec en Outaouais), and Eric Ruppert (York University); the award will be presented at PODC 2026.

Danupon Nanongkai is a Scientific Director at the Max Planck Institute for Informatics in Saarbrücken, Germany since 2022, where he heads the Algorithm and Complexity department. He received a Ph.D. in Algorithms, Combinatorics, and Optimization (ACO) from Georgia Tech in 2011 and a docent (aka habilitation) in Computer Science from KTH Royal Institute of Technology, Sweden, in 2017.

His research focuses on graph algorithms and computational complexity, with particular interest in developing algorithmic techniques that are effective across a range of computational models. He has contributed to efficient algorithms for fundamental graph problems, such as connectivity and distances, in settings ranging from sequential to distributed and dynamic algorithms.

Further information:
Prize Announcement by PODC/DISC: https://www.podc.org/2026-edsger-w-dijkstra-prize-in-distributed-computing/
Website of the Algorithms and Complexity department: https://www.mpi-inf.mpg.de/de/departments/algorithms-complexity

Editor:
Philipp Zapf-Schramm
Max Planck Institute for Informatics
Phone: +49 681 9325 4509
Email: pzs@mpi-inf.mpg.de

The Saarland Ministry of Economic Affairs, in cooperation with the Saarland Informatics Campus, has presented the Journalism Prize for Informatics. The German Informatics Society is a partner of the prize. The main prizes were awarded to a radio feature by Austrian Broadcasting (ORF), an article from the magazine “c’t – Magazin für Computertechnik” , and an online piece by Swiss Radio and Television (SRF). A special prize was awarded to an article in the children’s science magazine “GEOlino.”

The Journalism Prize for Informatics was first announced in 2006 and is endowed with a total of 16,000 euros. The prize money for the main awards in the three categories “Text,” “Audio,” and “Video and Multimedia,” each endowed with 5,000 euros is provided by the Saarland Ministry of Economic Affairs, Innovation, Digitalization, and Energy. Additionally, the German Research Center for Artificial Intelligence (DFKI) is once again sponsoring a special prize endowed with 1,000 euros this year. The goal of the Journalism Prize for Informatics is to promote high-quality reporting on informatics topics beyond specialist circles.

Jürgen Barke, Saarland’s Minister of Economic Affairs, Innovation, Digitalization, and Energy and patron of the prize, says: “With the Journalism Prize for Informatics, we honor outstanding and critical reporting on digital topics while strenghtening trust in complex technologies. Journalists make the opportunities and risks of informatics understandable, promote digital education, and enable citizens to form their own well-founded opinions—a central contribution to social responsibility in the age of artificial intelligence.”

This year, the jury evaluated a total of 84 submissions. The jury of the Journalism Prize for Informatics consists of Peter Bylda, a long-time editor of the Saarbrücker Zeitung and now a freelance journalist; Peter Hergersberg, editor-in-chief of the science magazine MaxPlanckForschung; Isabel Münch, Fellow of the German Informatics Society; Dr. Wolfgang Pohl, Managing Director of the nationwide informatics competitions; Florian Possinger from Saarländischer Rundfunk; Vera Sikes, Head of Department at the Federal Office for Information Security and Head of the BSI Saarbrücken office; Dr. Christel Weins, a scientist and co-founder of the journalism prize; freelance technology and science journalist Peter Welchering; and Prof. em. Dr. Dr. h.c. Reinhard Wilhelm, Professor of Informatics at Saarland University and long-time Director of the Leibniz Center for Informatics at Schloss Dagstuhl.

The main prize in the “Text” category, endowed with 5,000 euros, goes to Greta Friedrich for her article “Koste es, was es wolle: Big-Tech-Unternehmen verschwenden mit ihren Investitionen Ressourcen,” published on September 10, 2025, in the magazine “c’t – Magazin für Computertechnik” and on heise online. The article is available at:

https://www.heise.de/ratgeber/Wie-Big-Tech-Unternehmen-Umwelt-und-Mensch-gefaehrden-10508672.html

Jury’s Statement: In her article, Greta Friedrich highlights the devastating consequences of the AI boom for the environment and people. Using numerous examples and studies, she describes how major tech corporations recklessly consume resources in their “gigantomania” to stay ahead in the race for AI market leadership. The piece reports on overloaded power grids and (re)commissioned nuclear power plants supplying energy to data centers. It also addresses water shortages in regions surrounding these centers, as vast amounts are diverted to cool AI accelerators. Additionally, the article examines the psychological strain on workers fine-tuning language models. It provides an overview of the current state of affairs and compellingly argues that every use of AI should be questioned. The piece holds tech corporations accountable while raising awareness among users. An outstanding contribution to a socially explosive and highly relevant topic.

The main prize in the “Video & Multimedia” category, endowed with 5,000 euros, goes to Julian Schmidli, Pascal Albisser, Keto Schumacher, and Marina Kunz for their online article “Der toxische Sog der Manosphere,” published online on December 4, 2025, by Swiss Radio and Television (SRF). The piece is available at: https://www.srf.ch/news/schweiz/radikalisierung-auf-tiktok-der-toxische-sog-der-manosphere

Jury’s Statement: The visually sophisticated online feature by Julian Schmidli, Pascal Albisser, Keto Schumacher, and Marina Kunz powerfully demonstrates how quickly young people can be drawn into the so-called “Manosphere” on TikTok—a social media subculture defined by toxic masculinity, misogyny, and self-loathing. The piece reveals a disturbing reality: Within just five minutes of scrolling, the team’s test accounts encountered the first toxic content from this environment. A dangerously escalating trend with measurable impacts on the mental health of young men, as the authors substantiate with research. Through its multimedia approach—featuring striking graphics, embedded videos, and illustrations—the contribution explores multiple facets of the issue while creating its own compelling narrative pull, making it nearly impossible to look away. Especially relevant in the context of current debates on potential social media bans for minors, this is a highly insightful and urgent piece of journalism.

The main prize in the “Audio” category, endowed with 5,000 euros, goes to Sarah Kriesche for her radio feature „Wie Algorithmen unser Leben formen“, the fourth installment of a series on algorithms, broadcast on May 8, 2025, on Austrian Broadcasting Corporation (ORF) Ö1. The episode is available online at: https://oe1.orf.at/programm/20250508/794317/Wie-Algorithmen-unser-Leben-formen-4

Jury’s Statement: In a society increasingly shaped by algorithms, Sarah Kriesche explores a critical question: Which decisions should machines be allowed to make, and which must remain the domain of humans? The piece opens with a striking example from the Netherlands, where an AI system falsely accused people of welfare fraud for years—immediately establishing a strong sense of the stakes involved. From there, the contribution examines multiple dimensions of an algorithm-driven society: It explains the fundamentals of machine learning, addresses the problematic anthropomorphization of AI (the attribution of human traits and emotions), and tackles key questions of accountability, power, and transparency. It also considers how to empower citizens to engage critically with these systems. MAde accessible by fitting statements and musical interludes, the piece delivers an engaging listening experience while covering a breadth of topics. The jury particularly commends the meticulousness and accuracy with which this complex subject matter was handled.

The €1,000 Special Prize from the German Research Center for Artificial Intelligence (DFKI) is awarded to David Krenz for his article “So klappt’s mit der KI,” published on September 12, 2025, in GEOlino, the science magazine for children.

Jury’s Statement: In GEOlino, David Krenz takes on the challenging task of making complex computer science topics accessible to young readers. Using clear, age-appropriate language and engaging visuals, he explains how AI systems work, highlights practical applications, and addresses potential risks. Carefully selected experts help contextualize the content, while firsthand accounts from young AI users create relatable connections to children’s everyday lives. The jury particularly emphasizes the importance of introducing children to new technologies early and fostering a reflective, informed approach. David Krenz achieves this exceptionally well in his three-part series on AI, with the award-winning first installment standing out as a prime example of how to engage young audiences with cutting-edge topics.

Background Saarland Informatics Campus

To support this work, Höller will receive funding of up to €1.8 million through the German Research Foundation’s (DFG) Emmy Noether Programme.

The following text has been machine translated from the German with no human editing.

When planning complex processes, it is advantageous if the systems used can react flexibly to changes in the environment. ‘The advantage of systems from automated planning is that they can solve different problems without the system itself needing to be changed. They operate based on a model, a simplified description of the problem,’ explains Daniel Höller, an research associate at Saarland University. The behaviour of these model-based systems is mathematically provable and therefore reliable. It is also possible to explain exactly why a particular behaviour occurs. ‘However, these systems also have some disadvantages. In particular,  they may solve small problems but require a great deal of computing time once the problems become larger. Furthermore, the model must also be adapted to minor changes, making it relatively inflexible,’ explains Daniel Höller.

 For this reason, researchers are increasingly using machine learning algorithms in automated planning. Such computer programmes have the advantage that they can adapt flexibly and are more scalable. They can therefore be trained on small-scale scenarios and then also work for large-scale applications. ‘Compared to systems from automated planning, however, behaviour based purely on machine learning is difficult to interpret or explain to a human. Why, for example, does the robot do what it does?’ says the computer scientist, who has a doctoral degree in Computer Science.

 His aim is therefore to develop planning systems that combine traditional, explainable techniques with machine learning. ‘This combination is useful not only when solving planning problems, but also right from the model-building stage. If, for example, you look at a city’s road network and want to predict how long a particular journey will take, factors such as the day of the week, the time of day or weather conditions come into play. Using machine learning techniques, these factors can be incorporated into the model and thus into the planning process," explains Daniel Höller.

 A main application of machine learning in the context of automated planning is  to speed up the systems and find solutions more quickly. ‘Here, we will be working in particular on systems that can provide guarantees despite the integration of machine learning, such as the optimality of the resulting behaviour,’ explains the Computer Science expert. In addition to this integration of machine learning techniques into planning systems, work is also being carried out in the opposite direction. ‘We want to use techniques from automated planning to prove that following a learned action policy can never result in unsafe states,’ explains Daniel Höller, who has previously been researching the fundamentals of AI in Professor Jörg Hoffmann’s group at Saarland University. 

Daniel Höller’s research project on ‘Neuro-Symbolic Methods in Sequential Decision Making’ was successful in a special call for proposals on ‘Artificial Intelligence Methods’ under the Emmy Noether Programme. An internationally renowned panel of experts selected 36 out of 178 outline project proposals for submission of a full application. Of these, only 15 were selected for funding. Daniel Höller will now receive – subject to a successful interim evaluation – a total of 1.8 million euros to establish an Emmy Noether Group. Through the Emmy Noether Programme, the German Research Foundation (DFG) supports exceptionally qualified researchers in the early stages of their careers whose doctoral degrees were awarded no more than four years ago, who have international experience and who have completed a postdoctoral phase. 

Further information:

Emmy Noether Programme of the German Research Foundation (DFG): 

Press release on the call for proposals ‘Methods of Artificial Intelligence’

Emmy Noether Programme website 

Daniel Höller’s website:  https://fai.cs.uni-saarland.de/hoeller/

For further information, please contact:

Dr. Daniel Höller
Academic Research Associate
Foundations of Artificial Intelligence (FAI) Group
Email: hoeller(at)cs.uni-saarland.de

This new phase of the CRN focuses on understanding the heterogeneity of Parkinson’s disease and why it varies from person to person. To this end, research will be advanced to enable more precise diagnostics and the development of improved therapies. Novel resources will be created to enable the global research community to work on a shared, high-quality basis and to overcome technical barriers that currently impede drug development.

For over a decade, Andreas Keller, Professor of Bioinformatics at Saarland University, and his team have been researching neurodegenerative diseases such as Parkinson’s and Alzheimer’s. Their work focuses on microRNAs, which are short, non-coding segments of ribonucleic acid (RNA) that regulate the translation of genetic information within cells. ‘Analysing them generates huge amounts of sequencing data, which we process using our bioinformatics methods. As part of the newly approved funding, we will have access to over a hundred terabytes of data from Parkinson’s patients,’ explains Andreas Keller. Using artificial intelligence, his team will analyse this data to find novel RNA-based therapeutic candidates for the treatment of Parkinson’s disease.

‘Based on our analyses, we will propose RNA candidates suitable for therapeutic development. Our research partners at the Weizmann Institute in Israel and at Columbia University in the USA will then test these in cellular models. If the results are promising, we will investigate them further in preclinical studies with a view to progressing as quickly as possible towards clinical applications,’ says bioinformatician Andreas Keller, who also heads a research group at the Helmholtz Institute for Pharmaceutical Research Saarland.

Parkinson’s disease is shaped by multiple interacting factors. Biological characteristics such as age and gender influence the risk of developing the disease and its progression, yet they have often not been given sufficient consideration in molecular research to date. ‘At the same time, cell type, genetic background, and environmental influences contribute to its complexity. They also affect disease progression and how patients respond to treatment,’ explains Keller. Even within a single cell type, Parkinson’s disease can exhibit different molecular patterns influenced by these factors. ‘We expect to identify patterns within these large-scale datasets that will enable us to develop more effective and personalized drug therapies. We are already benefiting today from large datasets containing information on people with Parkinson’s disease, including blood samples, cerebrospinal fluid analyses and brain tissue data from deceased patients,’ he adds.

Using AI-supported methods developed over many years, the bioinformatics team in Saarbrücken will analyse these Parkinson’s datasets and correlate them with information on age groups and gender. The goal is to create a comprehensive, freely accessible molecular knowledge base for Parkinson’s disease, which will assist further research within the newly funded network. Key partners include Tal Iram and her team at the Weizmann Institute in Tel Aviv, as well as Philip De Jager, Vilas Menon and their respective teams at Columbia University in New York. The US$9 million grant (approximately €7.7 million) from ASAP will be awarded primarily to Saarland University.

Further information: 

https://www.asapcrn.org/funding

https://www.ccb.uni-saarland.de/
https://www.helmholtz-hips.de/de/forschung/people/person/prof-dr-andreas-keller/

 Press photographs that can be used free of charge with this press release can be found at the bottom of the following web page. 

For further information, please contact: 
Prof. Dr. Andreas Keller
Tel. +49 681 302 68611
Email: andreas.keller(at)ccb.uni-saarland.de

Karol Myszkowski is honored with the Eurographics Outstanding Technical Contributions Awardfor his long-standing and impactful work in computer graphics research. The laudation notes that his work “[…] for decades has helped shape how we understand, model, and visualize the world,” and that his contributions have made him one of the most relevant researchers in the field in Europe. His research bridges the gap between graphics and perception “with exceptional elegance and technical soundness, defining standards of excellence in many fields,” particularly in high dynamic range imaging (HDR), a technique for representing images with especially high contrast and detail. Among other things, Karol Myszkowski is co-author of the book High Dynamic Range Imaging: Acquisition, Display, and Image-Based Lighting, which has since become a standard reference for students, researchers, and practitioners alike.

Furthermore, he is well-known for his work on image quality analyses and metrics. He has also made fundamental contributions to so-called “perceptual rendering,” developing methods of image generation guided by human perception that restrict detail reconstruction to what is perceptually discernible.

Other important research areas in which he has made important contributions include stereoscopic imaging, where a slightly different image is generated for each eye, creating 3D depth in films or VR applications, as well as the realistic rendering of materials. More recently, he has increasingly focused on virtual reality research. Overall, his work has helped to use computational resources more efficiently.

In addition to his scientific work, Myszkowski is deeply involved in the professional community. In 2020, he chaired the Technical Papers Program of SIGGRAPH Asia and is active in numerous committees and editorial boards. Particularly noteworthy is his commitment to supporting early-career researchers, where he is regarded as a role model through his mentorship and scientific integrity.

Karol Myszkowski has been conducting research at the Max Planck Institute for Informatics since 2000, where he leads the research group “HDR Imaging, Rendering and Advanced Displays” as a senior researcher. Prior to that, he worked for many years in Japan, including as an associate professor at the University of Aizu. His academic career began at the Szczecin University of Technology. He received his PhD in 1991, followed by his habilitation in 2001, both in computer science at the Warsaw University of Technology. In 2011, he was awarded a lifetime professor title by the President of the Republic of Poland. In 2025, he was already named a Eurographics Fellow and inducted into the ACM SIGGRAPH Academy, the latter being among the highest international honors in computer graphics research.

The “Young Researcher Award” for Marc Habermann underscores the role of the Max Planck Institute for Informatics in promoting internationally leading research in computer graphics.

The award recognizes Habermann’s significant contributions to human performance capture, digital human reconstruction, and photorealistic human rendering. His work innovatively combines computer graphics, computer vision, and machine learning, as highlighted in the laudation.

During his doctoral research, Marc Habermann introduced “LiveCap,” the first method capable of generating a detailed, movable 3D model of a human, including clothing deformations, from image data captured by a standard camera, such as those found in smartphones. He subsequently developed “DeepCap,” the first learning-based framework for high-fidelity human performance capture, for which he received a CVPR Best-Paper Honorable Mention. Currently, Marc Habermann is researching how classical computer graphics methods can be combined with machine learning approaches (neural-explicit methods) to create even more realistic and efficient digital humans. Such technologies could be used in the future for virtual meetings, games, films, and anywhere digital humans play a role.

Habermann’s research findings have been published in the most important conferences and leading journals in computer graphics and computer vision, including SIGGRAPH, ACM Transactions on Graphics, CVPR, ICCV, ECCV, EUROGRAPHICS, and NeurIPS. His work has received multiple awards, including the EUROGRAPHICS PhD Award, the DAGM MVTec Dissertation Award, and the Otto Hahn Medal from the Max Planck Society.

Since 2017, the Saarland native has been conducting research at the Max Planck Institute for Informatics, initially as a doctoral student in the group of Professor Christian Theobalt. After earning his doctorate in November 2021, he began leading the “Graphics and Vision for Digital Humans Group” and serving as the Scientific Manager of the Real Virtual Lab in the Visual Computing and Artificial Intelligence department headed by Director Christian Theobalt. In December 2024, he was appointed Senior Researcher at the Institute.

Further Information:
Karol Myszkowski’s Website: https://people.mpi-inf.mpg.de/~karol/

Award citation for Karol Myszkowski: https://www.eg.org/wp/eurographics-awards-programme/the-outstanding-technical-contributions-award/outstanding-technical-contributions-award-2026-karol-myszkowski/

Website of the Graphics and Vision for Digital Humans group: https://gvdh.mpi-inf.mpg.de/

Award citation for Marc Habermann: https://www.eg.org/wp/eurographics-awards-programme/the-young-researcher-award/young-researcher-award-2026-marc-habermann/

Website of the Real Virtual Lab: https://www.mpi-inf.mpg.de/de/departments/visual-computing-and-artificial-intelligence/real-virtual-lab

Website of the Eurographics Conference 2026: https://eg2026.github.io/

Editor:
Philipp Zapf-Schramm
Max Planck Institute for Informatics
Phone: +49 681 9325 4509
Email: pzs@mpi-inf.mpg.de

This is an incredible achievement — only two out of over 5,000 accepted ICLR papers have received such an award this year!

MPI-SWS papers:

1. Automata on S-adic Words. Valérie Berthé, Toghrul Karimov, and Mihir Vahanwala (ICALP, Track B)
2. Hypersequent calculi have Ackermannian upper bounds. A. R. Balasubramanian, Vitor Greati and Revantha Ramanayake (LICS)
3. Infinite-state games with energy objectives beyond counters. Irmak Saglam and Georg Zetzsche (ICALP, Track B)
4. On the Subspace Orbit Problem and the Simultaneous Skolem Problem. Piotr Bacik and Anton Varonka (LICS)
5. On Variable-Bounded Non-Linear Expansions of Presburger Arithmetic. Piotr Bacik, Joris Nieuwveld, Joël Ouaknine, Mihir Vahanwala, Madhavan Venkatesh and Emil Rugaard Wieser (LICS)
6. Optimally Controlling a Random Population. Hugo Gimbert, Corto Mascle, Patrick Totzke (ICALP, Track B)
7. Optimal Sequential Flows. Hugo Gimbert, Corto Mascle, Patrick Totzke (ICALP, Track A)
8. Population Protocols over Ordered Agents. Michael Blondin, Michaël Cadilhac, Benjamin Courchesne, Lucie Guillou, Corto Mascle, and Isa Vialard (ICALP, Track B)
9. The Complexity of Nested Reset Counter Systems. A. R. Balasubramanian and Franzisco Schmidt (LICS)
10. The complexity of downward closures of indexed languages. Richard Mandel, Corto Mascle and Georg Zetzsche (LICS)

MPI-SP papers:

1. Complete Relational Logic for Infinite-Dimensional Quantum Programs with Unbounded Assertions. Gilles Barthe, Minbo Gao, Jam Kabeer Ali Khan, Matthijs Muis, Ivan Renison, Keiya Sakabe, Michael Walter, Yingte Xu, Tianshi Yu and Li Zhou (LICS)

MPI-INF papers:

1. A Faster Directed Single-Source Shortest Path Algorithm. Ran Duan, Xiao Mao, Xinkai Shu, Longhui Yin (ICALP, Track A)
2. Computing the (k+2)-Edge-Connected Components in k-Edge-Connected Digraphs in Subquadratic Time. Loukas Georgiadis, Evangelos Kipouridis, Evangelos Kosinas, Charis Papadopoulos, Nikos Parotsidis (ICALP, Track A)
3. Faster algorithms for k-Orthogonal Vectors in low dimension. Anita Dürr, Evangelos Kipouridis, Michael Lampis, Karol Wegrzycki (ICALP, Track A)
4. Fast decremental tree sums in forests. Benjamin Aram Berendsohn, Marek Sokołowski (ICALP, Track A)
5. Improved Tree Sparsifiers in Near-Linear Time. Daniel Agassy, Dani Dorfman, Haim Kaplan (ICALP, Track A)
6. Low Rank MSO. Mikołaj Bojańczyk, Michał Pilipczuk, Wojciech Przybyszewski, Marek Sokołowski and Giannos Stamoulis (LICS)
7. Node-Weighted Triangles: Faster and Simpler. Shyan Akmal, Nick Fischer (ICALP, Track A)
8. Permutation Patterns in Streams. Benjamin Aram Berendsohn (ICALP, Track A)
9. Random Access in Grammar-Compressed Strings: Optimal Trade-Offs in Almost All Parameter Regimes. Anouk Duyster, Tomasz Kociumaka (ICALP, Track A)

The DFKI Robotics Innovation Center in Bremen has a long-standing and successful partnership with SENAI CIMATEC. As early as 2013, SENAI (Serviço Nacional de Aprendizagem Industrial) opened the “Brazilian Institute of Robotics” (BIR) in Salvador, Bahia – modeled after the DFKI – with Prof. Dr. Dr. h.c. Frank Kirchner serving as Scientific Director. SENAI is the central organization for industrial training, applied research, and innovation in Brazil. Among other things, this collaboration resulted in the autonomous underwater vehicle (AUV) FlatFish for inspecting offshore energy infrastructure. The framework agreement now signed aims to further deepen this partnership, particularly in the areas of applied AI, robotics, and industrial innovation.

With the NEXUS AI Campus, DFKI also gains a new strategic partner. Located in Joinville, Santa Catarina, in Brazil’s industrially strong southern region, this innovation hub focuses on applied AI research, technological development, and the implementation of market-oriented solutions. NEXUS connects companies, universities, and international centers of excellence and specifically promotes global collaboration and knowledge transfer. Key areas of focus include generative AI, large language models, computer vision, health tech, industrial AI, quantum computing, cognitive robotics, and applications in agriculture and smart cities.

The goal of both partnerships is to jointly advance the development and application of innovative AI technologies, intensify the exchange of knowledge and talent, and provide new impetus for the economic use of AI.

“With Brazil as the partner country, this year’s Hannover Messe offers a particularly fitting setting to showcase the further development of our collaborations. Intensifying our collaboration with partners such as SENAI CIMATEC and NEXUS AI Campus opens up additional opportunities for us to translate our European AI research perspective and expertise into industrial applications and to further expand our international network with like-minded partners,” said Prof. Dr. Antonio Krüger, CEO of DFKI.

DFKI at Hannover Messe 2026

Hannover Messe is regarded as the world’s leading platform for technologies related to industrial transformation and brings together thousands of companies from the mechanical engineering, electrical, and digital industries, as well as the energy sector, every year. This year, DFKI will be presenting practical AI solutions in Hall 11, Booth B30, as well as at partner booths – ranging from robotic systems for aerospace and logistics to assistance solutions in medicine, management, and law, and the optimization of complex processes in industry and healthcare. Here, visitors can experience the transfer from research to industrial application firsthand, while also seeing how the developed technologies can be successfully applied to other fields.

How can artificial intelligence be effectively transitioned from research to real-world use? At Hannover Messe 2026, the German Research Center for Artificial Intelligence (DFKI) will demonstrate how AI solutions can be deployed in an economically viable, socially responsible, and technologically autonomous way. From April 20 to 24, DFKI will showcase practical AI technologies and transfer methods in Hall 11, Booth B30, as well as at partner booths, thereby strengthening Germany as an innovation hub and bridging scientific excellence and practical application.

The DFKI will present modern AI approaches applicable across various domains. Intelligent robotic systems support resource extraction on the Moon and simulate weightlessness, proving equally effective in logistics for industrial goods. As assistance and decision-support tools, AI technologies demonstrate their utility in diverse fields: from medical diagnostics and therapies to business decision-making and legal contract review. Additionally, AI systems model and optimize complex processes – such as those in medium-sized manufacturing firms or hospital emergency rooms – contributing significantly to increased efficiency, transparency, and quality assurance.

“There has never been a better time to invest in industrial and European AI. The necessary infrastructure is being built in Europe, and we have the ideas. Anyone who isn’t yet using AI to enhance products and processes should do so now, or risk losing their competitive edge. In the areas of Trusted AI, space, and healthcare, we are already experiencing the value AI provides.” — Professor Antonio Krüger, Scientific Director and CEO of DFKI

The strategic innovation fields of Trusted AI, Health AI, and Space AI define DFKI’s main themes at Hannover Messe

Trusted AI stands for transparent, secure, and compliant AI systems throughout the entire lifecycle – from data collection to practical application. Its goal is to build trust as the foundation for creating economic value and to actively help shape European standards. Trusted AI ensures that AI systems can be verified, robust, and securely certified – for autonomous systems, critical infrastructure, and industrial processes. DFKI develops methods and standards that combine security, explainability, and sovereignty.

With Health AI, DFKI integrates cutting-edge AI methods into medical practice – from diagnostics and personalized therapy to intelligent assistance systems in care and rehabilitation. DFKI combines image processing, signal analysis, and clinical expertise – for more efficient patient care and life-saving pattern recognition.

In space exploration, AI is reaching its full potential: robotics, real-time data analysis, and mission control. Space AI opens new horizons for autonomous systems, satellite data utilization, and resilient infrastructure. These advancements not only strengthen Europe’s role in space exploration but also boost industry, environmental monitoring, and supply systems on Earth. DFKI solutions for Earth observation and extraterrestrial exploration can be easily adapted to address challenges on Earth.

In the stage program at Hannover Messe, DFKI researchers will discuss key future topics such as AI gigafactories and the path to technological sovereignty, agentic AI and resilient factories, trustworthy AI as the foundation of Europe’s digital sovereignty, mobile robots for extraterrestrial use, physical AI in rehabilitation, and industrial AI in a transnational context.

This honor is granted to only about 3.6% of the more than 1,700 papers accepted at the Conference on Human Factors in Computing Systems. The conference takesplace in Barcelona from April 13 to 17, 2026.

The first award-winning paper is “Scene2Hap: Generating Scene-Wide Haptics for VR from Scene Context with Multimodal LLMs.” Paul Strohmeeier explains: “In virtual reality, we are used to seeing or hearing content. Virtual worlds you can feel by touch are much rarer. While visual content is created through light and acoustic content through sound waves, our approach is based on vibration. Building on simple effects like those one might recognize from VR controllers or smartphones, we recreate the complex dynamics of the tactile world this way.” However, the vibration patterns required to create such haptic impressions (vibrotactile feedback) currently still have to be created manually, which does not scale for complex VR scenes with many objects. That is the focus of the newly awarded paper.

With “Scene2Hap,” first authors Arata Jingu from Professor Jürgen Steimle’s Human-Computer Interaction Lab at Saarland University and Easa AliAbbasi from Paul Strohmeier’s Sensorimotor Interaction group have now developed an approach for automatically designing meaningful vibrotactile feedback for objects and scenes in virtual reality. To do this, the researchers use a multimodal large language model (LLM) that can process not only language, but also image and audio data. The model automatically infers the semantics of objects, physical properties and material characteristics, as well as the physical context of the scene. “We draw on various layers of meta-information, ranging from the context of a virtual object to material properties that the LLM can recognize in the image,” explains Easa AliAbbasi. The vibrotactile feedback is then generated and transmitted separately to each hand via the VR controllers being held. In three different user studies, the team was able to show that “Scene2Hap” successfully improved users’ sense of space and perception of materials and generally contributed to a better user experience when the VR environment was created entirely with the newly developed pipeline.

The second award-winning paper, “How are Vibrotactile Experiences Visually Represented? A Taxonomy of Illustration Characteristics,” is a meta-study that examines how haptic impressions and tactile information are communicated in research. More specifically, the award-winning paper investigates how vibrotactile feedback is represented visually. “When new methods for visual rendering are developed, their quality can be shown directly in papers, for example through an image. In haptics research, things are different: we can only describe what something feels like, but we cannot easily convey the actual sensation directly. In my opinion, this limited ability to represent haptic experiences is a central challenge in haptics research,” says Paul Strohmeier.

To analyze this issue, the researchers first developed a taxonomy for representing vibrotactile experiences (VTX) and then collected a total of 1,652 papers from the past 25 years from the digital libraries of the two world’s largest professional associations in computing, the Association for Computing Machinery (ACM) and the Institute of Electrical and Electronics Engineers (IEEE). Within these, they identified 768 visual representations from 409 research papers and coded them according to their visual representation of VTX based on the taxonomy. Their results indicate that (1) half of the illustrations communicate on the timing of vibrotactile feedback with regards to users’ actions, (2) illustrations depict stimuli rather than experiences and infrequently communicate multimodal aspects of the experiences, and (3) contextual information of vibrotactile displays and experiential aspects are often distributed across several complementary figures.

“With our taxonomy, we want to give authors a tool to analyze and improve their illustrations. At the same time, in combination with the corresponding dataset it could serve as an approach for generative models to automatically create ideas or inspiration for visualizing one’s own research,” explains Dennis Wittchen from the SensInt group, who co-authored the paper as first author together with Bruno Fruchard from the French research institute Inria.

Orignal publications:
Arata Jingu, Easa AliAbbasi, Sara Safaee, Paul Strohmeier, and Jürgen Steimle. 2026. Scene2Hap: Generating Scene-Wide Haptics for VR from Scene Context with Multimodal LLMs. In Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI ’26), April 13–17, 2026, Barcelona, Spain. ACM, New York, NY, USA, 21 pages. https://doi.org/10.1145/3772318.3791297

Bruno Fruchard, Dennis Wittchen, Nihar Sabnis, Paul Strohmeier, and Donald Degraen. 2026. How are Vibrotactile Experiences Visually Represented? A Taxonomy of Illustration Characteristics. In Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems (CHI ’26), April 13–17, 2026, Barcelona, Spain. ACM, New York, NY, USA, 24 pages.
doi.org/10.1145/3772318.3790598

Further information:
Website of the Conference on Human Factors in Computing Systems: https://chi2026.acm.org/
Website of the Sensorimotor Interaction group: https://sensint.mpi-inf.mpg.de/

Editor:
Philipp Zapf-Schramm
Max Planck Institute for Informatics
Phone: +49 681 9325 4509
Email: pzs@mpi-inf.mpg.de