Reliable Conversational Domain-specific
Data Exploration and Analysis
January 12-17, 2026 | Verona, Italy
The ARMADA Winter School brings together 15 doctoral candidates and supervisors from 8 partner institutions across Europe for an intensive week of training, research presentations, and networking.
For questions or comments, please contact the project coordinator: Matteo Lissandrini at project.armada@ateneo.univr.it
The Feedback Loop of Intelligence: Operationalizing User Signals in Frontier Model Alignment
Explored how production telemetry and implicit user signals drive the development of frontier LLMs, examining the engineering pipelines that transform noisy behaviors into reward models while addressing challenges like sycophancy and reward hacking.
Neuro-Symbolic Integration: Reasoning, Guarantees and Explainability
Introduced the latest advancements in Neuro-Symbolic AI, focusing on building models with reliability guarantees and addressing challenges in trustworthiness, robustness, and interpretability for safety-critical scenarios.
Designing Transparent, Fair, and Reliable AI Systems: Beyond Architecture and Metrics
Examined an integrated approach to trustworthy AI where causal reasoning, domain constraints, and value-sensitive design are embedded throughout the model development lifecycle, with practical examples of operationalizing auditability and equity.
Geographic and Geocultural Biases in LLMs
Investigated how geographic and geocultural biases arise in large language models through interactive workshop sessions, examining methods for measuring and mitigating bias while co-creating ideas for new evaluation tasks and benchmarks.
Dates: January 12-17, 2026
Location: Verona, Italy
Participants: 15 Doctoral Candidates from 8 partner institutions and local PhD students