I agree. This whole "papers written and reviewed by people" is so anachronistic! I say let #NeurIPS2026 accept papers in "latent tokens"--which are evaluated by the reviewer LLMs.. (Once we know humans are not writing, we don't need humans reviewing, and as long as we know humans are not reviewing the papers, may as well go with latent tokens--which are more expressive, right? Such "Quantum Papers" can lead to quantum leaps in Science..🙄)

@JessicaHullman @dhadfieldmenell These authors explicitly agreed to a policy that they either did not use AI or disclosed AI at submission time. In this case, authors explicitly agreed not to do type C and then did.

@jennajrussell @dhadfieldmenell Yes thank you, I should have noted that when mentioning that example. Will add

@JessicaHullman Type C becomes Type B when the operator loses justificatory lineage. Type C becomes Type A when the operator maintains full derivational control (which you self-typed in your other post) 😅