LLM reviewers draw 50% of weaknesses from the appendix vs 8% for humans

plopesresearch · x · 2026-11-18

Shared research data shows LLM reviewers source 50% of their listed weaknesses from a paper's appendix, compared to just 8% for human reviewers. The accompanying chart also reveals markedly different attention distributions across main-body pages. The general trend: human reviewers focus on the big picture, overall presentation, and support for major claims, while AI reviewers nitpick details — a key consideration for AI-assisted peer review.

Related event: Study: LLM Reviewers Fixate on Appendices Far More Than Humans(4 posts)→

Original post →

More from Models

Models channel →