Research forbetter AI.

    We research how AI can work directly with knowledge, reason through logic and return answers that can be verified.

    Topics

    Featured research

    Preprint paperBilAI
    2026

    Agentic SPARQL: Evaluating SPARQL-MCP-powered Intelligent Agents on the Federated KGQA Benchmark

    Daniel Dobriy, Frederik Bauer et al.

    Book chapterPublished
    2026

    Future Innovations in Combining Large Language Models and Knowledge Graphs

    Daniel Dobriy

    Conference paperAccepted/In press
    2026

    Layer-Wise Localization and Classification of Hallucinations in Large Language Models

    Naveen Lamba, Sanju Tiwari et al.

    Publications

    2026
    Conference paperAccepted/In press

    Agentic AI, Context Engineering and Knowledge Graphs: Current Approaches, Challenges and Opportunities

    Niraj Karki, Manjila Pandey, Sanju Tiwari, Nandana Mihindukulasooriya, Sven Groppe, Daniel Dobriy

    With the recent advancements in Large Language Models (LLMs) and Agentic AI, Context Engineering (CE) has emerged as a novel research area. CE aims to fill the prompts for LLM Agents with relevant contextual knowledge required to perform complex tasks, where the quality of this context is paramount for reliability. Knowledge Graphs (KGs) offer a promising approach to integrate diverse contextual knowledge based on Semantic Web and Knowledge Representation approaches. In this paper, we study current approaches to identify challenges and opportunities for utilising KGs in CE and explore their limitations and strategic future research directions. The findings illustrate inconsistencies in methodologies and limited understanding of scalability and quality assurance challenges, which slow down the development of robust, context-aware AI systems capable of dealing with real-world complexity and multi-domain reasoning tasks.

    Conference paper/Context Engineering/Knowledge Graphs/Large Language Models/Ontology/Knowledge Representation/Quality
    Open paper
    2026
    Preprint paperBilAI

    Agentic SPARQL: Evaluating SPARQL-MCP-powered Intelligent Agents on the Federated KGQA Benchmark

    Daniel Dobriy, Frederik Bauer, Amr Azzam, Debayan Banerjee, Axel Polleres

    DOI: 10.57938/83c86964-2d48-46f1-b655-5bef78c1a837

    Standard protocols such as the Model Context Protocol (MCP) that allow LLMs to connect to tools have recently boosted "agentic" AI applications, which, powered by LLMs' planning capabilities, promise to solve complex tasks with the access of external tools and data sources. In this context, publicly available SPARQL endpoints offer a natural connection to combine various data sources through MCP by (a) implementing a standardised protocol and query language, (b) standardised metadata formats, and (c) the native capability to federate queries. In the present paper, we explore the potential of SPARQL-MCP-based intelligent agents to facilitate federated SPARQL querying: firstly, we discuss how to extend an existing Knowledge Graph Question Answering benchmark towards agentic federated Knowledge Graph Question Answering (FKGQA); secondly, we implement and evaluate the ability of integrating SPARQL federation with LLM agents via MCP (incl. endpoint discovery/source selection, schema exploration, and query formulation), comparing different architectural options against the extended benchmark. Our work complements and extends prior work on automated SPARQL query federation towards fruitful combinations with agentic AI.

    Artificial intelligence/Knowledge engineering/Web engineering/Semantic technologies/Working Papers on Information Systems, Information Business and Operations
    Open paper PDF
    2026
    Book chapterPublished

    Future Innovations in Combining Large Language Models and Knowledge Graphs

    Daniel Dobriy

    This chapter examines how large language models and knowledge graphs can be integrated to support factual grounding, traceable reasoning, and persistent, updatable knowledge. It develops a taxonomy covering graph-grounded retrieval and generation, KG construction, access and maintenance, and hybrid adaptive architectures. It reviews evaluation resources, emphasizing the need to assess graph access, evidence retrieval, updating, provenance, uncertainty, memory stability, and computational cost separately. Particular attention is given to graph-based agent memory, where KGs support persistent knowledge, world modeling, inference, tool and agent coordination, and governance of shared assertions. A running example of a scientific literature analyst illustrates these capabilities and their operational requirements. The chapter identifies key challenges in graph versioning, runtime logical validation, executable graph access, source discovery, and metadata interoperability. It outlines three future directions: data search engines combining natural-language interaction with federated querying over the decentralized Web of Data, a common semantic layer connecting diverse knowledge organization systems, and generative models that operate directly on language-independent semantic representations. It argues that the value of integration depends on task requirements and on maintaining graph memory that is accurate, auditable, reversible, and governable over time.

    Book chapter/Knowledge graphs/Large language models/Neuro-symbolic AI/Agentic AI/Graph-based agent memory/GraphRAG/Semantic Web
    Open paper
    2026
    Conference paperAccepted/In press

    Layer-Wise Localization and Classification of Hallucinations in Large Language Models

    Naveen Lamba, Sanju Tiwari, Manas Gaur, Daniel Dobriy

    LLM/AI/Hallucinations/Explainability
    Open paper
    2026
    Conference paperPublished

    On the Historic Roots of Agentic AI in Semantic Web Services

    Axel Polleres, Frederik Bauer, Daniil Dobriy, Tobias Käfer, Lukas Kubelka, Andreas Harth, Thomas Wehr

    DOI: 10.1145/3774905.3795076

    Conference paper/WWW Companion '26: Companion Proceedings of the ACM Web Conference 2026 : Dubai
    Open paper PDF
    2025
    Journal paperPublished

    Be Open or Competitive? Exploring the differences in the adoption of (open) data sharing in European railway public transport

    Shahrom Sohi, Sri Harikrishnan, Amin Anjomshoaa, Axel Polleres

    DOI: 10.1016/j.trip.2025.101698

    Open data shape the interactions between governments, businesses, and the public. In the European railway sector, the push towards liberalisation, environmental sustainability, and a Single European Railway Area (SERA) has intensified demands for open, interoperable public transport data. Yet, railway organisations face a dual challenge: fulfilling their public service role in contributing to the transport and mobility open data ecosystem while safeguarding organisation’s sensitive and competitive information. This paper investigates how national railway organisations across the European Union and the European Free Trade Association (EFTA) participate in open data initiatives, focusing on three channels: Railway Open Data Portals (ROPs), Governmental Open Data Portals (GOPs), and National Access Points (NAPs). Through a structured data collection and manual labelling approach, this study catalogues 1097 open datasets across national railway operators and infrastructure managers, categorising them by portal type, country, thematic focus, and quality based on the 5-Star Open Data Model. The results reveal a fragmented Open Data landscape of the European railway, characterised by disparities in dataset volume, thematic focus, and openness standards. Operational data lead publications such as timetables, schedules and routes, but significant gaps persist in areas such as customer experience, specific infrastructure and organisation data. Railway operators prefer ROPs and GOPs over EU mandatory mobility portals NAPs for publishing open data, raising interest about consistency and compliance with European legislative frameworks. Beyond openness, the uneven availability of railway open data across the Union raises questions of digital equity. As unequal access to transport information risks reinforcing informational disparities among EU and EFTA residents, resulting a call for action for harmonised open data policies and stronger quality controls at the European level.

    Journal paper/Transportation Research Interdisciplinary Perspectives
    Open paper PDF