@nazneenrajani Super interesting work, Nazneen et al.! Quick question: what kinds of bugs do you think it is possible for LMs to generate natural language descriptions for currently? I see that this is currently based on topics, but could it be extended to say linguistic phenomena, or author characteristics?