In this age of information “abundance”, there are duelling challenges: for authors, to stand out and be noticed; for readers, to sort and digest the endless stream of content.
To that end, readers have begun uploading publications to LLMs for summarization and evaluation. Thus, authors – in order to stand out – have taken to embellishing their work in order to gain prominence from the rest of the chaff. And, by embellishing, I mean they’ve inserted specific prompts and other hidden content in the papers by making certain text “white”, or using infinitesimal font sizes:
The idea here is to have content practically invisible to the reader, but digested by the LLM.
Cheeky! And, per report, not necessarily effective. ChatGPT apparently responded to the additional prompts, but Claude and Gemini did not.