• KG construction, completion, and refinement with GenAI and Agents
  
  • Text-to-KG extraction using LLMs (multilingual, multimodal)
  • KG completion, cleaning, and refinement (deduplication, entity resolution)
  • Fact verification, contradiction detection, and consistency checking
  • Human-in-the-loop KG curation and interactive refinement
• KG grounding for information retrieval, including generation, querying, and dialogue
• KG-grounded generation / GraphRAG (subgraph retrieval, path-based evidence)
• Natural language querying of KGs via GenAI (e.g., NL-to-SPARQL)
• Hybrid QA and dialogue systems combining KGs and GenAI (e.g., Agents)
• Prompting / controllable generation using KG structure and constraints
• Context and memory indexing and retrieval for GenAI and agents
• Neuro-symbolic methods, reasoning, and explainability
• Hybrid reasoning with rules, constraints, and structured evidence
• Explainability and verifiable reasoning with provenance/evidence graphs
• Cross-domain knowledge transfer with KGs and GenAIK
• Representations, embeddings, and temporal/evolving KGs
• GenAI for KG embeddings and hybrid vectorgraph representations
• Temporal KGs, dynamic updates, continual learning, concept drift
• Ontology learning, schema induction, alignment, and schema evolution
• Trustworthiness, safety, and governance
• Bias mitigation using KGs in GenAI and Agentic Systems
• Hallucination reduction via grounding/constraints; robustness to attacks
• Uncertainty estimation and calibrated confidence
• Privacy, access control, and policy-aware KG-grounded generation
• Agentic KGs and real-world systems
• Agentic KGs: KGs as long-term memory/state for LLM agents
• KGs serving agents' memories: Episodic (experiences, events, etc.), Semantic (facts, concepts, etc.), and Procedural (skills, tasks, etc.)
• KG-aware planning, tool use (query/update), and multi-agent coordination
• Collaborative & shared agent memories and contexts.
• Context Engineering enhanced by KGs
• GenAI/Agents and KG Applications
• Efficient and proactive personal assistance & Personalisation
• Multi-Lingual & Multi-modal integrations and enablement
• Personalisation vs Generalisation in GenAI and Agentic systems memory
• Domain-specific applications: scholarly knowledge, biomedicine & healthcare, finance, education, etc.
• Task-specific applications: personal assistance, dialogue systems, recommender systems, customer service, etc.
• Architectures and pipelines
• Datasets, benchmarks, and evaluation
• Benchmark datasets for GenAI or Agents plus KG tasks
• Evaluation of grounding/faithfulness, factuality, reasoning, robustness
• Evaluation pitfalls: data leakage, LLM-as-a-judge bias, reproducibility, and reporting standards

I. Policies

General Information: The workshop is archivable, and its proceedings will be published in CEUR-WS. All accepted and presented papers will appear in the workshop proceedings. At least one of the authors of the accepted papers must register for the workshop and present their submission(s) to be included in the workshop proceedings. The workshop will be a 100% in-person 1-day event at IJCAI-ECAI 2026. Use of Generative AI Policy: This workshop adheres to the CEUR-WS GenAI Policy. Please familiarise yourself with the policy and include a Declaration on Generative AI section in your manuscript. Dual Submission Policy: Dual submissions, such as submitting the same manuscript to more than one venue (i.e. workshop, conference, journal, etc.), are not allowed. Withdrawal Policy: Failing to register will lead to a paper withdrawal.

II. Submission Tracks

There are three types of submissions covering the entire joint-workshop topics spectrum (see above):

• Research papers: Full (8-10 pages), Short (max 6 pages): presenting novel research or work in progress.
• Industry papers (max. 6 pages): in which industry experts can present and discuss practical solutions, use cases, and best practices at any stage of implementation.
• Position & demo papers (max. 4 pages): encouraging papers describing significant work in progress, late-breaking results or ideas, as well as functional systems relevant to the community.

These page limits only apply to the main body of the paper. Authors may include an unlimited but reasonable number of pages of references and appendices. In addition, papers must include a mandatory section on the limitations of the work, placed after the conclusions and before the references. Authors may also optionally include a section discussing ethical considerations and concerns regarding their research. Each submission must be submitted to only one track, the most suitable one.

III. Submission Format

All submissions must follow CEUR-ART single-column format. An official Overleaf template is available for LaTeX users. The offline templates can be downloaded here CEURART.zip, which contains the CEURART style and also the ODT (LibreOffice) template file. More information about the CEURART templates is available in the CEURART style files for papers section on the CEUR-WS page.

Each submission shall be One Single PDF file, including the references and appendices.
Supplementary materials (of reasonable number/size) may be included, but they are optional, and reviewers are not required to review these materials.
Submissions that do not adhere to the specified styles, including paper size, font size restrictions, and margin width, will be desk-rejected.
The reviewing process will be single anonymous, wherein each paper will be reviewed by at least three Program Committee members. A meta-review will be provided in case of any disagreements. The final decision of acceptance/rejection will be made in consensus by the Chairs.

IV. Submission Link

Papers should be submitted to OpenReview. Note: Please be aware of OpenReview's moderation policy when creating new profiles without an institutional email.

V. Presentation

All accepted papers shall be presented in person. Accepted papers will be presented as posters only or as posters plus oral presentation. Authors required to deliver an oral presentation will be notified accordingly.