Artificial Intelligence (AI) is increasingly embedded in critical systems, in various domain such as automotive, railways, aerospace, healthcare, defense to manufacturing where performance, safety, and robustness are mandatory. While AI-driven systems demonstrate remarkable capabilities, their deployment in high-risk environments introduce complex challenges: unpredictable failures, adversarial vulnerabilities, ethical dilemmas, and the need for rigorous validation across dynamic operational conditions.

Ensuring the reliability, resilience, and trustworthiness of AI-based systems requires interdisciplinary collaboration among researchers, engineers, regulators, and domain experts. This special track seeks novel contributions that advance the state-of-the-art in measuring, modeling, verifying, and guaranteeing the performance, safety, and robustness of AI systems, spanning theoretical foundations, methodological innovations, and real-world deployments.

We welcome contributions that address (but are not limited to) the following key themes:
Foundations of Robust and Safe AI
- Formal methods for verification, validation, and certification of AI systems (e.g., machine learning, logic-based, symbolic AI, hybrid AI, generative AI, agentic, GPAI…).
- Uncertainty quantification and probabilistic guarantees for AI decision under ambiguity.
- Adversarial robustness and cybersecurity of an AI system: Defenses against evasion, poisoning, robbery and distribution shifts in training/inference.
- Fault tolerance and resilience: Mechanisms for graceful degradation and recovery in AI critical systems.
- Explainability and interpretability in safety-critical contexts, including causal reasoning and counterfactual analysis.

Performance Assurance and Benchmarking
- Metrology for AI: Standardized metrics and benchmarks for safety-critical performance (e.g., latency, accuracy, fairness, reliability, …).
- Testing and edge-case evaluation: Systematic approaches to identify failure modes in AI constituents and AI systems.
- Real-time monitoring and runtime assurance: Techniques for continuous validation of AI behavior in operational environments, and for continuous safety assessment.
- Performance-safety trade-offs: Balancing efficiency with robustness in resource-constrained systems.

Safety-Critical AI Engineering
- Design principles for safe AI: Architectures that enforce safety principles.
- Human-AI collaboration: Ensuring safe interaction between autonomous systems and human operators.
- Safety cases and argumentation frameworks for AI-based systems (e.g., compliance with ISO 21448, IEC 61508, ARP6983 or other domain-specific standards).
- Regulatory and compliance challenges: Aligning AI systems with evolving safety and ethical standards (e.g., AI Act).

Agent design for explainable behaviors
- System autonomy and architectures
- Argumentation techniques, conflict resolution
- Agent-based negotiation
- Multi-agent and swarm robustness: Coordination and fault tolerance in decentralized AI systems Reinforcement learning and collaborative policy

Robustness in Dynamic and Uncertain Environments
- Generalization and adaptability: Ensuring robustness across distribution shifts, domain gaps, and long-tail scenarios.
- Lifelong learning and continuous validation: Methods for updating AI models without compromising safety.
- Planning under uncertainty, probabilistic techniques, belief states exploration, POMDP
- Physics/Geometric-informed Neural Network, neuro-symbolic and hybrid AI: Combining various AI technics with first-principles knowledge for safety.

Societal and Ethical Dimensions
- Bias, fairness, and accountability in high-stakes AI applications.
- Transparency and auditability: Tools for regulators, insurers, and end-users to assess AI system trustworthiness.
- Risk communication and user trust: Bridging the gap between technical guarantees and public perception. Responsible use.
- Case studies and lessons learned from deployments in automotive, aerospace, healthcare, railways, or defense.
- Appropriation, acceptability and learning curves with trustworthiness

Tools, Frameworks, and Industrial Applications
- Open-source tools for robustness testing, verification, or safety monitoring.
- Industry use cases: Real-world deployments of robust AI in automotive, aerospace, finance, or critical infrastructure.
- Standardization efforts: Contributions to emerging norms (e.g., CEN-CENELEC, ISO…) for AI safety.

These are only suggestions; papers discussing other issues related to Performance, Safety and Robustness in AI-based systems are welcome.

Important Dates
- Submission: April 04, 2026
- Notification: April 26, 2026
- Registration: May 4, 2026
- Camera-ready: May 10, 2026

Contribution Types
- Regular papers [in the proceedings, digital library] from 6 pages to 8 pages
- Short papers (work in progress) [in the proceedings, digital library] from 4 pages to 6 pages
- Posters: two pages [in the proceedings, digital library]
- Posters: slide only [slide-deck posted on www.iaria.org]
- Presentations: slide only [slide-deck posted on www.iaria.org]

Paper Format
- See: http://www.iaria.org/format.html
- Before submission, please check and comply with the editorial rules: http://www.iaria.org/editorialrules.html
Publications
- Extended versions of selected papers will be published in IARIA Journals: http://www.iariajournals.org
- Print proceedings will be available via Curran Associates, Inc.: http://www.proceedings.com/9769.html
- Articles will be archived in the free access ThinkMind Digital Library: http://www.thinkmind.org

Papers Submission
- https://www.iariasubmit.org/conferences/submit/newcontribution.php?event=PESARO+2026+Special
- Please select Track Preference as PSRAI
Registration
- Each accepted paper needs at least one full registration, before the camera-ready manuscript can be included in the proceedings.
- Registration fees are available at http://www.iaria.org/registration.html