Values in the wild: Discovering values in real-world language model interactions

Are the conversations(&#x2F;“conversations”) in this study using a model that is too big to apply their circuits analysis to?<p>It would be interesting if we could see these “values” in terms of how the different features interact with each-other.