Efthymia Tsamoura - Eight Years of Weakly Supervised Neurosymbolic Learning: Main Results and Algorithms

Neurosymbolic learning (NSL)—the integration of neural and symbolic mechanisms for inference and learning—has been proposed as a remedy for some of the most critical limitations of neural networks. A problem that has received considerable attention in the relevant literature is that of weakly supervised learning in the presence of a symbolic component. In this problem, one or more neural classifiers perform symbolic grounding, mapping low-level inputs onto high-level symbolic concepts. A symbolic component then infers outputs consistent with both the predicted concepts and any provided prior knowledge, and learning is driven by supervision applied only to the outputs of the symbolic component. Research has uncovered an intriguing pitfall. In particular, symbol grounding can "deceive" the learning process: even when the neural classifiers produce incorrectly grounded concepts, the symbolic component may still infer outputs that match the ground truth.

This talk provides an overview of research on weakly supervised NSL over the last eight years, with a focus on the above challenge. We begin by discussing early learning frameworks, then present the main theoretical results—including PAC-learnability, reasoning shortcuts, and learning imbalances. We conclude with an overview of strategies for improving learning efficiency, delving into a specific technique that enhances the accuracy of learned classifiers by up to 53% through better exploiting the representation space.

Asim Munawar - Neuro-Symbolic AI for Reliable LLM Agents

LLM agents are powerful, but they still struggle with reliable reasoning, planning, and consistency in real-world workflows. This talk explores how neuro-symbolic AI can improve agent reliability by combining neural models with symbolic reasoning, verification, structured memory, and tool-based execution. The presentation will cover emerging architectures for multi-step reasoning agents, practical enterprise challenges, and directions toward more trustworthy and explainable AI systems.

• Enhancing Retrieval-Augmented Generators with Symbolic Query Definitions, Marco Calamo, Laura Papi, Marco Console, Massimo Mecella, Filippo Bianchini

• A Fair Objective for Human-Empowerment-Preserving AI: Desiderata, Design, and Likely Behavioral Consequences, Jobst Heitzig, Ram Potham

• SymStep: Symbolic Step Verification for Logical Reasoning, Aida Usmanova, Rui Gao, Dilshod Azizov, Ricardo Usbeck, Zangir Iklassov

• Accelerating Verification for Neuro-symbolic Multi-agent Systems via Compact MILP encodings, Elena Botoeva, Cosmo De Bonis-Campbell, Francesco Leofante, Panagiotis Kouvaros

• Blowup-Blowdown World Models for LLM Agents: Symbolic Quotients and Counterexample-Guided Refinements Do Not Commute, Hikaru Matsuoka

• Towards a New, Semi-Explainable Machine Learning Architecture Using Dimensionality Reduction and Neural-Inspired Grammatical Evolution, Jakub Skrzyński, Antoni Ligęza

• From Formalisers to Engineers: An Agentic Position to HTN Domain Model Generation, Ilche Georgievski, Ebaa Alnazer

Invited Speaker [TBA]

Contributed Talks [10 mins + 2 mins questions each]

• From Exploration to Reuse: An Embodied Agent Framework for Manipulation Skill Learning, Mohamed Roshdi, Alexander Zorn, Jörg Kindermann, Hermann Blum. 

• Scalable Neurosymbolic LLM Reasoning via Multi-Token Encoding, Varun Dhanraj, Chris Eliasmith

• Improving Sequential Decision-Making in LLM Agents via Experience Memory, Jakub Rada, Viliam Lisý

• RML Meets LLMs: More Structure, Fewer Errors, Shikhat Karkee, Elena Botoeva, Anna Jordanous, Davide Lanti

Giuseppe Marra - Deep Probabilistic Logics: from statistical relational to neurosymbolic AI.

This talk presents a unified formal perspective on neurosymbolic methods, which combine learning and reasoning in AI, showing how many seemingly different approaches can be understood through the lens of deep probabilistic logics. It further illustrates how this perspective can provide a foundational semantic substrate for emulating existing approaches and developing the next generation of neurosymbolic methods.

Luis Lamb - [TBA]

[TBA]

• The Epistemically Blind OS: A Case for Neurosymbolic Operating Systems, Fawaz Ishola

• TIMELY-Agent: Privacy-Aware Neuro-Symbolic Orchestration for Temporal Clinical Benchmark Construction, Linglong Qian, Ao Zhang, Haoyu Wang, Zitong Li, Zina Ibrahim

• Neurosymbolic Scene Graph Generation in the Open-Vocabulary Setting, Lukas Arzoumanidis, Jannik Matijevic, Youness Dehbi

• Aligning LLMs Beyond the West: Culturally Generalizable Safety for Women’s Cancer Advice in Lingala, Patrick Tenga Shako, Mohamed Hamed Kholief, Mahmoud El-Alem

• Even LLMs deserve a body: How grounded chess representations impact language-mediated strategic reasoning, Sebastiano Tocci, Marco Roveri

• Automated Extraction of Symbolic Abstractions in Production and Materials Engineering, Mustafa Awd

• Neuro-Symbolic Agents for Regulated Process Automation: Challenges and a Research Agenda, Alexander Rombach, Chantale Lauer, Nijat Mehdiyev

Contributed Talks [10 mins + 2 mins questions each]

• Neurosymbolic Optical Character Recognition, Quinten Dewulf, Robin Manhaeve, Wannes Meert, Luc De Raedt.

• Behavioral Competence Without Conceptual Structure: Probing Type Knowledge in a Neural Pokémon World Model, Roger Zhu

• MeSH-SNOMED-15K: A Heterogeneous Biomedical Entity Alignment Benchmark, Vaibhava Lakshmi Ravideshik, Mayank Kejriwal

Discussion Panel - Including All Invited Speakers and Q&A Session 

Closing