RAIDO NOVEMBER 2025 NEWSLETTER.

Welcome to the Raido Project’s November newsletter! This issue summarizes our key activities, research breakthroughs, and event participation during the autumn months of September and October 2025.

SUMMARY

The autumn of 2025 has been a period of validation and high-impact dissemination for the RAIDO project. The most significant highlight of this period was the successful completion of our project review meeting in September, where the consortium received positive remarks from reviewers, confirming that our trajectory toward trustworthy and resource-efficient AI is on target.

Simultaneously, the project has seen a surge in academic output, with the submission and presentation of four major research papers addressing critical areas such as Alzheimer’s detection, fairness in ML, IoT architecture, and privacy-friendly image retrieval. Coupled with our active participation in global forums, from Florence to Macau, RAIDO continues to solidify its presence in the international AI research landscape.

PROJECT COORDINATION & REVIEW

In September 2025, the RAIDO consortium gathered for a critical review meeting. We are pleased to report that the meeting was highly successful. The project received positive feedback on the work completed so far, validating both our technical progress and our administrative coordination. This successful review serves as a green light for the next phase of RAIDO, energizing our partners to continue their excellent work in developing decentralized and trustworthy AI frameworks.

RESEARCH EXCELLENCE & PUBLICATIONS

This quarter has been exceptionally productive regarding technical contributions. RAIDO partners have submitted four significant papers to top-tier conferences and journals, showcasing our commitment to “Ethics by Design” and technical innovation:

• “NeuroXVocal: Detection and Explanation of Alzheimer’s Disease through Non-invasive Analysis of Picture-prompted Speech” – RAIDO researchers presented “NeuroXVocal,” the first end-to-end explainable Alzheimer’s Disease classification system. This novel architecture combines a multimodal classifier with a RAG-based explainer to provide clinically interpretable explanations. Achieving 95.77% accuracy on benchmark datasets, this tool bridges the gap between AI predictions and clinical decision-making. Available [here]
• “Fairbeat: Assessing and Mitigating Bias with the Composite Balance Score” – Addressing the critical challenge of fairness, we introduced “Fairbeat,” a tool that leverages the Composite Balance Score (CBS) to assess dataset balance before model training. This tool allows developers to visualize and mitigate bias regarding protected attributes, promoting more equitable AI systems. Watch the demo: https://youtu.be/9aHKfZgtXKg and access the full paper [here]
• “Resource Efficient Federated LoRaWAN Architecture for Far-Edge IoT Applications” – Focusing on Edge AI, this study proposes a “Federated LoRaWAN (LoRA-FL)” system. By combining privacy-preserving Federated Learning with Knowledge Distillation, this architecture solves strict bandwidth and energy limitations in rural IoT deployments (such as smart agriculture), proving that scalable intelligence is possible even in far-edge environments. Full paper [here]
• “Vision-Free Retrieval: Rethinking Multimodal Search with Textual Scene Descriptions” – Challenging standard Vision-Language Models, this paper introduces a privacy-friendly “text-to-text” retrieval paradigm. By using VLLM-generated structured image descriptions instead of raw images, the model reduces the modality gap and achieves state-of-the-art zero-shot performance with models as small as 0.3B parameters, offering a more privacy-centric alternative for search tasks. Full paper [here]

RECENT EVENTS & DISSEMINATION

September and October 2025 were marked by a series of high-profile engagements where RAIDO shared its expertise with the global community:

IEEE CH Workshop on Green Trustworthy AI (Sept 2025) – RAIDO made a strong showing at the IEEE-CH conference in Florence, specifically within the GTAI workshop. The project contributed two papers and one presentation, highlighting our work at the intersection of cultural heritage and trustworthy AI. Read more [here]

AI, Data and Robotics Forum (ADRF) (Sept 2025) – In late September, RAIDO representatives attended the ADRF in Stavanger, Norway. This forum allowed us to engage with European policymakers and industry leaders, discussing how RAIDO’s contributions support a resilient and strategically autonomous European AI ecosystem. [link]

UNU AI Summit 2025 – Expanding our global reach, RAIDO participated in the United Nations University (UNU) AI Summit in Macau. This event provided a unique platform to discuss the global implications of our research in trustworthy AI. Read more [here]

ICCV 2025 Workshop (Oct 2025) – In October, RAIDO successfully supported the ICCV 2025 workshop on “Binary and Extreme Quantization.” This collaboration underscored our leadership in energy-efficient Deep Learning, focusing on low-bit quantization techniques that reduce computational power while maintaining model accuracy. Read more [here]

STAY CONNECTED

To read more about these updates and stay current with RAIDO’s latest deliverables, publications, and events, please visit our news page: https://raido-project.eu/news/

RAIDO aims to make research always publicly available. You can find all of our publications in our dedicated Zenodo Community.

FIND US ON SOCIAL MEDIA