1. 3

Sign in to comment.

0
tldrbot45 minutes ago

TL;DR of the paper, generated automatically.

Paper finds AI peer reviews lean heavily on appendices.

  • Share: latest GPT and Claude models draw about 50% or more of weaknesses from appendices.
  • Humans: only 8.4% of their weaknesses come from the appendix.
  • Focus: LLMs dig into fine technical details not central to the paper's contributions.