Plenary Speakers

Large language models have demonstrated remarkable generative and reasoning capabilities, yet their reliability remains limited by hallucination, weak grounding in external knowledge, and the difficulty of assigning credit to individual reasoning steps. This talk presents a progression of approaches toward more faithful and trustworthy reasoning with language models, moving from explicit structural constraints to self-evolution, learned evaluation, and process-supervised reinforcement learning. We first ground reasoning in structured evidence through knowledge graph (KG)-constrained decoding, ensuring that generated reasoning follows valid and connected facts. We then introduce Generate, Discriminate, Evolve (GenDiE), which generates and scores candidate sentences and iteratively improves context-faithful generation. Next, Retrieval-Augmented Generation Zero Evaluator (RAG-Zeval) enables transparent evaluation by decomposing responses into claims, grounding them in evidence, and learning to rank response quality through reinforcement learning (RL). Finally, Tree-based Process Supervision for Reinforcement Learning in Agentic Retrieval-Augmented Generation (TREEPS-RAG) learns from intermediate reasoning and retrieval decisions through step-wise credit assignment, advancing agentic RAG beyond sparse final-outcome supervision. Together, these approaches point toward more trustworthy and capable AI systems that reason over evidence, generate faithfully, evaluate their outputs transparently, and continuously improve their reasoning and retrieval strategies through process-supervised learning.

Professor Helen Meng is Patrick Huen Wing Ming Chair Professor of Systems Engineering and Engineering Management at The Chinese University of Hong Kong (CUHK). She is the Founding Director of CUHK’s Ministry of Education Key Laboratory in Human-centric Computing and also CUHK’s InnoCentre on AI, Centre for Perceptual and Interactive Intelligence. Her team’s AI-enabled, spoken langage-based dementia screening project was awarded the first HKSARG RGC Theme-based Research Scheme award in AI in 2019. As Curriculum Lead of the CUHK-JC AI4Future Project since 2019, she developed Hong Kong’s first pre-tertiary AI curriculum, emphasizing ethics, earning the 2021 Hong Kong ICT Gold Award. The abridged version of the curriculum has also been adopted by the HKSAR Government’s Education Bureau, to be taught in all local secondary schools starting in 2024. Helen’s accolades include the IEEE SPS Leo Beranek Meritorious Service Award (2019) for leadership and service, and fellowships from the IEEE, International Speech Communication Association, Hong Kong Institution of Engineers and Distinguished Fellow of the Hong Kong Computer Society. Her team has secured numerous awards, including First Prize at the DialDoc@ACL 2022 Challenge, INTERSPEECH Best Student Paper Awards (2023, 2024), US National Academy of Medicine Healthy Longevity Global Grand Challenge, and gold, silver and bronze medals from the International Exhibition of Inventions Geneva (2024, 2025).

The contemporary digital information ecosystem is under serious strain. What researchers have come to call Information Disorder, a phenomenon broken down into misinformation, disinformation, and malinformation by Wardle and Derakhshan (2017), is no longer a marginal problem: it is a structural threat to democratic societies. Scholars across disciplines have documented how the deliberate spread of false or misleading content erodes institutional trust, distorts electoral processes, and fuels social polarisation (Lewandowsky et al., 2023; Sunstein, 2017). The European External Action Service (EEAS) has warned that coordinated Foreign Information Manipulation and Interference (FIMI) campaigns represent one of the most serious threats to European democracy, while the World Economic Forum ranked misinformation and disinformation as the top global risk over a two-year horizon in its 2024 Global Risks Report. The situation has worsened sharply with the mass adoption of generative AI: manipulative content is now cheaper to produce, harder to detect, and capable of reaching millions of people within hours.

Against this backdrop, the SERICS (Security and Rights in Cyberspace) initiative has funded the Information Disorder Awareness (IDA) project, which develops AI-based methodologies for identifying and analysing cognitive manipulation. IDA combines multi- agent credibility assessment with Explainable AI (XAI) techniques to improve transparency and accountability in automated decision-support systems, with the broader ambition of evolving current monitoring platforms into an infrastructure capable of protecting citizens, institutions, and democratic processes at scale.

Current research increasingly points to multi-agent systems built on Large Language Models (LLMs) as a promising approach to these challenges. Networks of LLM agents can reason collaboratively, assess credibility, and reach consensus, making disinformation detection more robust and easier to interpret. Semantic and ontological knowledge representations are also being explored to anchor AI analyses in structured knowledge. More speculatively, some work combines quantum computing principles with neuro- symbolic prompting to simulate debunking strategies in advance, evaluating their likely effectiveness before real-world deployment.

Taken together, these developments point toward a future in which hybrid human–AI teams will be central to information analysis and decision-making. Preserving information integrity in such environments is no longer just a matter of spotting deceptive content: it requires frameworks that support cognitive resilience, trustworthy knowledge construction, and informed collective decision-making. At its core, this is a question of whether public trust and social cohesion can survive, and be strengthened, in societies increasingly mediated by AI.

Vincenzo Loia obtained a degree in Information Sciences from the University of Salerno in 1985, and a PhD in Computer Science in 1989 from the Université Pierre & Marie Curie Paris VI, France. He is currently serving as full professor of computer science at the University of Salerno, where he held the position of researcher from 1989 to 2000, and as associate professor from 2000 to 2004. From the beginning of his career he has shown a deep research interest in the use of innovative approaches for solving complex problems through new software design and programming paradigms as agent and multi-agent systems; in the design and testing of systems for approximate reasoning, through the synergistic use of standard semantic technologies, description logic and Fuzzy Logic; in the definition of original conceptual data analysis methodologies for the extraction of knowledge from unstructured resources; in the creation of decision support systems conveyed by artificial intelligence models, cognitive architectures, computational intelligence and granular computing.

The pillars that characterize his scientific research over the years have found confirmation in various application domains, from the medical one to the legal one, and in the last ten years in the field of cybersecurity. Precisely in this last area he demonstrates his great predisposition for multidisciplinarity, coordinating the scientific committee of the „Multidisciplinary Observatory for the fight against organized crime and terrorism” which implements a continuous scientific research activity applied to the thematic areas of Computational Intelligence to combat terrorist activities spread on the Web and Deep Web, in support of information analysis projects for crimes under the jurisdiction of the DNAA (National Anti-Mafia and Anti-Terrorism Directorate ). DNAA is an Italian judicial body within the Prosecutor General’s Office at the Supreme Court of Cassation. It coordinates investigations into organized crime and terrorism, managing district anti-mafia directorates (DDA) nationwide to ensure cohesive prosecution. This experience allowed him to coordinate the project financed by the Presidency of the Council of Ministers for the implementation of a system that supports the Intelligence Analyst in searching for information on the web thanks to the collection, indexing and structuring of open sources. President of the SERICS Foundation (Security and rights in the cyber space https://serics.eu/ ) expression of the Extended Partnership (projects financed by the PNRR EU with 118 million of euro) on line 7, the one dedicated to cybersecurity, new technologies and protection of rights, of which the University of Salerno and national hub.

Finally, both in the planning field and in scientific research activities he has demonstrated notable team building skills by aggregating heterogeneous skills around original and innovative ideas. All this is demonstrated by the large number of co-authorships, on scientific publications, with colleagues from other scientific-disciplinary sectors and by the collaborations established in the context of research and development projects, with corporate partners and departments of scienti He has produced around 450 scientific works published in international conference proceedings, book chapters, international journals as well as monographs.

Included in Stanford University’s World’s Top 2% Scientists list through at 2025

 

Editorial experiences:

Editor-in-Chief and Founder of the international journal “Ambient Intelligence and Humanized Computing” (Cite score 11.8, IF 7.104), Springer-Verlag; (from the birth of this journal in 2009 to 2024).

Editor-in-Chief and Founder of the international book series “Big and Integrated Artificial Intelligence” Springer-Verlag (from 2020).

Assistant Editor, Managing Editor, Editor-in-Chief, Editor Responsible for Special Issues for international journal “Soft Computing” (Cite score 8.1, IF 2.5), Springer-Verlag (from 1990 to 2024).

Editor-in-Chief of the international journal “Evolutionary Intelligence”, (Cite score 7.4, IF 2.6),

Springer-Verlag (from 2018 to 2025);

Editor-in-Chief Journal of Uncertain Systems, (Cite score 2.4, SJR 0.422) World Scientific Connect (from 2021)

IEEE (biennial position taken during 1995-2005)

Associate Editor for IEEE Transactions on Systems, Man, and Cybernetics;

Associate Editor for IEEE Transactions on Industrial Informatics;

Associate Editor for IEEE Transactions on Fuzzy Systems;

Societies:

IEEE Senior Member

He has held numerous roles within the IEEE Computational Intelligence Society, including:

– Chair of Emergent Technologies Technical Committee (ETTC) – Chair of Task Force “Intelligent Agents”,

– Chair of Task Force “Ambient Intelligence”,

– Chair of IEEE CIS Italian Chapter,

– Vice-Chair IEEE Intelligent Systems Applications Technical Committee.

Institutional roles and activities: 

prof. Loia has held various institutional roles at the University of Salerno:

– Director of the Department of Mathematics and Computer Science (DMI) (from 2006 to 2009);

– Director of the Department of Computer Science (DI) (from 2012 to 2016);

– Director of the Department of Business Sciences – Management & Innovation Systems (DISA-MIS) (from 2015 to 2019).

– Coordinator of the Board of Department Directors (from 2014 to 2017);

– Member of the Academic Senate of the University of Salerno (from 2016 to 2019).

– Scientific Manager of the multidisciplinary Observatory for the fight against organized crime and terrorism, established between the University of Salerno and the National Anti-Mafia and Anti-Terrorism Directorate (DDNA) (from 2018 to 2022).

-President of CUR Campania (Regional Coordination Committee of Campania Universities) (from 2020 to 2023)

-Scientific manager for UNISA of the “Quantum Valley” project financed by the Campania Region with 100 million euros.

– Rector of the University of Salerno (from 2019 to 2025).

It is currently  President of the SERICS Foundation (from 2022) and Principal Investigator (PI) Spoke 2 „Disinformation & Fake News” Project PNRR SERICS. SERICS has been financed by Minister of Research with 118 million euros.

-Member of the Steering Committee and coordinator of the technical-scientific committee of the Teledife-Unisa Observatory for cybersecurity (from 2023).

– Member of IGF Italy Committee, established at the Department for Digital Transformation of the Presidency of the Council of Ministers (from 2024).

– Principal Investigator for the Dept Scienze Aziendali, Management e Innovation Systems (DISAMIS) of the project SINTESI (Sistemi Cyber-fisici Intelligenti e Sicuri per la Sanità, l’Industria e la Società) – PON RIC 27, founded with Euro 1.660.000

Teaching activity:

During his career, prof. Loia has held numerous teaching positions in various degree courses at the University of Salerno, including Fundamentals of Computer Science, Agent Systems, Semantic Web, Methodologies and Tools for It Project/Service Management, Safe Systems, Situation Awareness Systems, Safe and Awareness Systems, Cognitive Defense Systems, Information Disorder and Global Security, cyber security and cyber intelligence techniques. He has also been invited as a Visiting Professor to degree courses, masters and training seminars at prestigious national and international universities.

To be uploaded.

Yi Zeng is a Wu Yuzhang Chair Professor and Ph.D. Supervisor at the Gaoling School of Artificial Intelligence, Renmin University of China. He also serves as the Founding Dean of Beijing Institute of AI Safety and Governance (Beijing-AISI), and Director of the Beijing Key Laboratory of Safe AI and Superalignment. He is the Chair of the Mind Computing Technical Committee of the Chinese Association for Artificial Intelligence (CAAI) and Co-Chair of the AI Committee of the World Internet Conference (WIC). Additionally, he is a member of the United Nations Advisory Body on AI and a member of the UNESCO Ad Hoc Expert Group on AI Ethics. His research focuses on AI Ethics, Safety and Governance, Brain-Inspired AI, and AI for Sustainable Development. He was named one of the TIME 100 Most Influential People in AI (TIME 100/AI, 2023).