Company perspective

Model changes, clearer evidence: starting with runway perception

Fionn Labs explores model-change assurance: understanding what changes when a model changes, with runway perception as a concrete research application.

Start with what changed

A new model version brings a practical question: what changed in its behavior, and what does that mean for the next engineering decision? An improvement in one measure does not answer every question about the change.

That is the principle behind Fionn Labs' interest in delta assurance. The focus is on the difference between model versions and the evidence needed to understand it: what improved, what may have regressed, and what remains uncertain.

Evidence that helps the next decision

Testing a model change should help engineers decide what to investigate next. A useful comparison makes the trade-offs and limits visible rather than reducing the decision to a single score.

Our research asks how that evidence can be made useful to the people developing and evaluating models. It is not a claim that every change can be approved, or that a component-level result establishes the safety of a whole system.

Runway perception is a starting point

Runway perception gives this broader question a concrete aviation context. It is a research application for examining model behavior and changes, not the boundary of the company's interest in model-change assurance.

Fionn Labs is at the research stage. We are not presenting a deployed landing system or claiming a flight-safety or certification outcome. Earlier notes remain available as historical context; their questions can still be relevant without implying a current product or validated capability.