Highlights on AI research results

The general idea of probabilistic relation modeling and lifted reasoning

Why PRM is central to AI?

• It mirrors the structure of the world – objects linked by relations, all observed through noisy lenses.
• It gives AI a principled way to share knowledge across entities, enabling data‑efficient learning and true generalization.
• It unifies two historically separate AI traditions (symbolic reasoning and statistical learning) under a single mathematical framework.
• It provides a natural substrate for causal, counterfactual, and explainable reasoning, which are essential for trustworthy AI.
• It fuels a thriving ecosystem of models, algorithms, and applications that already power many state‑of‑the‑art systems in NLP, vision, recommendation, science, and robotics.

For these reasons, probabilistic relational modeling is not just another niche technique—it is a foundational pillar of any AI system that aspires to operate in the messy, interconnected, and uncertain reality we live in.

For an introduction to the general research context see the book "An Introduction to Lifted Probabilistic Inference" edited by Guy Van den Broeck, Kristian Kersting, Sriraam Natarajan and David Poole.

Summary of the main ideas of lifted rrobabilistic relational representation and reasoing

Summary of the main ideas of temporal lifted rrobabilistic relational representation and reasoing

Data Stream Fusion and Interpretation

This work already started at the University of Lübeck.

New techniques for data stream fusion and interpretation are developed as part of the dissertion of Marisa Mohr.

Marisa Mohr (2022). Multivariate Ordinal Pattern Representations for Time Series: Learning from Ups and Downs. (Doctoral dissertation, Universität zu Lübeck).

LJT: Lifted Junction Tree Algorithm

This work already started at the University of Lübeck.

LJT provide new optimization techniques for lifted reasonigng in the case of mulitple queries and has been developed as part of the dissertion of Tanya Braun.

T. Braun, & R. Moller (2016). Lifted junction tree algorithm. In Proceedings KI 2016: Advances in Artificial Intelligence - 39th Annual German Conference on AI (pp. 30–42). Springer.

Tanya Braun (2020). Rescued from a Sea of Queries: Exact Inference in Probabilistic Relational Models. (Doctoral dissertation, Universität zu Lübeck).

DLJT: Taming exact inference in temporal probabilistic relational models

This work already started at the University of Lübeck.

DLJT provides new optimizations for answering multiple temporal queries and has been developed as part of the dissertion of Marcel Gehrke. DLJT also solves the problem that reasoning need to be kept lifted in case of evidence accumulated over time. This is done with approximation techniques, but with guaranteed error bound.

Marcel Gehrke (2021). Taming Exact Inference in Temporal Probabilistic Relational Models. (Doctoral dissertation, Universität zu Lübeck).

LDJT: Lifted Dynamic Junction Tree Algorithm

M. Gehrke, T. Braun, & R. Moller (2018). Lifted Dynamic Junction Tree Algorithm. In 23rd International Conference on Conceptual Structures, ICCS 2018 (pp. 55–69). Springer.

Denoising the Future: Top-p Distributions for Moving Through Time

Florian Marwitz, Ralf Moller, Magnus Bender, & Marcel Gehrke (2025). Denoising the Future: Top-p Distributions for Moving Through Time. In Symbolic and Quantitative Approaches to Reasoning with Uncertainty (pp. 118–132). Springer.

PETS: Predicting efficiently using temporal symmetries in temporal probabilistic graphical models

Florian Marwitz, Ralf Möller, & Marcel Gehrke (2025). PETS: Predicting efficiently using temporal symmetries in temporal probabilistic graphical models. International Journal of Approximate Reasoning, 179.

Assumptions, Defaults, and Reclustering in Dynamic Parametric Factor Graphs

This work already started at the University of Lübeck.

New techniques for data stream fusion and interpretation are developed as part of the dissertion of Nils Finke.

Nils Finke (2023). Navigate Through Troubled Waters. (Doctoral dissertation, Universität zu Lübeck).

The general idea

Detailed view

Reclustering

Dynamic Parametric Factor Graphs with Continuous Randvars, Gaussian Bayesian Networks, Gaussian Processes, Approximations

This work already started at the University of Lübeck.

New techniques for lifting with continuous random variables, for dealing with Gaussian Bayesian Networks and Gaussian Processes, and approximation techniques for query answering dissertion of Mattis Hartwig.

Mattis Hartwig (2022). New Methods for Efficient Query Answering in Gaussian Probabilistic Graphical Models. (Doctoral dissertation, Universität zu Lübeck).

Estimating Causal Effects in Partially Directed Parametric Causal Factor Graphs

Malte Luttermann, Tanya Braun, Ralf Möller, & Marcel Gehrke (2024). Estimating Causal Effects in Partially Directed Parametric Causal Factor Graphs. In Proceedings of the Sixteenth International Conference on Scalable Uncertainty Management (SUM-2024) (pp. 265–280). Springer.

Compression versus Accuracy: A Hierarchy of Lifted Models

Jan Speller, Malte Luttermann, Marcel Gehrke, & Tanya Braun (2025). Compression versus Accuracy: A Hierarchy of Lifted Models. In Proceedings of the Twenty-Eighth European Conference on Artificial Intelligence (ECAI-2025) (pp. 5051–5058). IOS Press.

Towards Explainability of Approximate Lifted Model Construction: A Geometric Perspective

Jan Speller, Malte Luttermann, Marcel Gehrke, & Tanya Braun (2025). Towards Explainability of Approximate Lifted Model Construction: A Geometric Perspective. In Proceedings of the Eleventh Workshop on Formal and Cognitive Reasoning (FCR-2025) (pp. 41–56). CEUR.

Approximate Lifted Model Construction

Malte Luttermann, Jan Speller, Marcel Gehrke, Tanya Braun, Ralf Möller, & Mattis Hartwig (2025). Approximate Lifted Model Construction. In Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence (IJCAI-2025) (pp. 9077–9085). IJCAI Organization.

Lifted Model Construction Without Normalisation: A Vectorised Approach to Exploit Symmetries in Factor Graphs

Malte Luttermann, Ralf Möller, & Marcel Gehrke (2025). Lifted Model Construction without Normalisation: A Vectorised Approach to Exploit Symmetries in Factor Graphs. In Proceedings of the Third Learning on Graphs Conference (LoG-2024) (pp. 46:1-46:17). PMLR.

Lifting Factor Graphs with Some Unknown Factors for New Individuals

Malte Luttermann, Ralf Möller, & Marcel Gehrke (2025). Lifting Factor Graphs with Some Unknown Factors for New Individuals. International Journal of Approximate Reasoning, 179, 109371.

Colour Passing Revisited: Lifted Model Construction with Commutative Factors

Malte Luttermann, Tanya Braun, Ralf Moller, & Marcel Gehrke (2024). Colour Passing Revisited: Lifted Model Construction with Commutative Factors. In Proceedings of the Thirty-Eighth AAAI Conference on Artificial Intelligence (AAAI-2024) (pp. 20500–20507). AAAI Press.

Planning the behavior of systems with 1000 agents and more

Tanya Braun, Marcel Gehrke, Florian Lau, & Ralf Moller (2022). Lifting in Multi-Agent Systems under Uncertainty. In Thirty-Eighth Conference on Uncertainty in Artificial Intelligence (UAI 2022) (pp. 233–243). PLMR.

Mechanisms for Data Sharing in Collaborative Causal Inference

Björn Filter, Ralf Möller, & Özgür Lütfü Özçep (2024). Mechanisms for Data Sharing in Collaborative Causal Inference. In KI 2024: Advances in Artificial Intelligence (pp. 86). Springer.

A Mechanism for Mutual Fairness in Allocating Replicable Resources

Björn Filter, Ralf Möller & Özgür Lütfü Özçep (2025). A Mechanism for Mutual Fairness in Cooperative Games with Replicable Resources. In Proceedings of the Twenty-Eighth European Conference on Artificial Intelligence (ECAI-2025) (pp. 1278-1285). IOS Press.

Fair Mechanisms for Replicable Resources: A General Approach Based on Analogical Beneficence

Björn Filter, Ralf Möller & Özgür Lütfü Özçep (2026). Fair Mechanisms for Replicable Resources: A General Approach Based on Analogical Beneficence. In PRIMA 2025: Principles and Practice of Multi-Agent Systems (pp. 348–365). Springer..

Medical Cause and Effects Analysis: Planning Treatments

Malte Luttermann, Edgar Baake, Juljan Bouchagiar, Benjamin Gebel, Philipp Gruning, Dilini Manikwadura, Franziska Schollemann, Elisa Teifke, Philipp Rostalski, & Ralf Möller (2024). Automated Computation of Therapies Using Failure Mode and Effects Analysis in the Medical Domain. German Journal of Artificial Intelligence.

Nano robotic agents

F.-L.A. Lau, T. Braun, R. Moller, & S. Fischer (2022). Using decPOMDPcoms to holistically model and program nanodevices and emergent nanonetworks. In NANOCOM '22: Proceedings of the 9th ACM International Conference on Nanoscale Computing and Communication (pp. 1–7). Association for Computing Machinery (ACM).

Tanya Braun, Marcel Gehrke, Florian Lau, & Ralf Moller (2022). Lifting in Multi-Agent Systems under Uncertainty. In Thirty-Eighth Conference on Uncertainty in Artificial Intelligence (UAI 2022) (pp. 233–243). PLMR.