Our workshop "LLM4XAI: Generative Models for Explainable AI Narratives" has been accepted at CIKM 2026

As AI systems are increasingly deployed in high-stakes domains, their decisions must be communicated in ways that are not only technically rigorous but also understandable to the people affected by them. Traditional Explainable AI (XAI) methods generate structured artifacts—such as feature attributions and counterfactual examples—but these outputs can remain difficult for non-technical users to interpret.

Large Language Models (LLMs) offer a promising opportunity to transform these artifacts into accessible natural-language explanations. However, using generative models for this purpose also introduces important challenges, including hallucination, loss of faithfulness, and misalignment between generated narratives and the actual behavior of the explained model.

💡 What is LLM4XAI?

LLM4XAI: Generative Models for Explainable AI Narratives is a new multidisciplinary workshop bringing together researchers and practitioners from Explainable AI, Natural Language Processing, Information Retrieval, Human–Computer Interaction, and Responsible AI.

The workshop investigates how generative and agentic systems can act as reliable mediators between complex technical explanations and end users, ensuring that XAI narratives remain grounded, accurate, useful, and accessible.

🔍 What will the workshop focus on?

The workshop will cover three main research directions:

1. Generative and Agentic Methods for XAI Narratives
   Approaches for transforming structured XAI artifacts into faithful and user-aligned natural-language explanations, including multimodal generation, narrative planning, causal grounding, multi-agent systems, and adaptive explanations.

2. Reliability, Faithfulness, and Evaluation
   Methods for evaluating factual consistency, detecting hallucinations, improving robustness, studying alignment with model reasoning, and developing reproducible benchmarks and human-centered evaluation protocols.

3. Personalized, Interactive, and Responsible XAI Systems
   Conversational and personalized explanation interfaces, tool-augmented and agentic XAI systems, human-in-the-loop decision support, and applications in high-stakes domains such as healthcare, finance, and public administration.

🤝 What to expect

LLM4XAI will be a half-day workshop co-located with CIKM 2026, featuring full and short paper presentations, invited talks, spotlight sessions, posters, and opportunities for interdisciplinary discussion and community building.

🌍 Why it matters

LLM4XAI aims to establish XAI narratives as a research topic in their own right, bridging the gap between the technical rigor of existing explanation methods and the accessibility required by real users. By bringing together academic and industrial perspectives, the workshop seeks to advance generative explanations that are not only fluent, but also faithful, responsible, and genuinely useful.

The workshop will take place on November 8, 2026, in Rome, Italy.

🔗 Visit the workshop website at the following link.