Summary
The workshop takes place at IJCAI-ECAI on 15 August and covers robustness across models, deployment and user interaction. Its agenda includes distribution shift, adversarial inputs, tool use, human interaction, monitoring and assurance.
Why it matters
Deployment failures increasingly arise from the interaction between models, tools and users rather than from isolated benchmark errors. Treating robustness as a lifecycle property is important for high-impact applications.
Related coverage
- AIM4DK frames disappearing knowledge as an AI reliability problem
- Agent security gets an attack-and-defence benchmark
- CAR-bench tests whether agents recognise their limits