# Inspect LogitTilt

> Steer a model's sampling toward a behaviour, so an eval surfaces examples of it
in far fewer samples than repeated sampling needs.


## LogitTilt

- [LogitTilt](https://adrskapars.github.io/inspect_logittilt/index.html.md): LogitTilt raises how often a model produces the behaviour an eval is looking for while keeping its output on-policy, so examples surface in far fewer samples than repeated sampling needs.