Show the judge a neighbour.
A small probe flags a sentence. An LLM judge dismisses it. One labelled example changes the verdict. What can that reversal teach us about explanations?
Read the post →Research notes
Blog
Experiments, ideas, and things I learned about AI along the way.