Researchers using Claude/ChatGPT for CVPR pre-review always get "borderline"

Prior_Brilliant_9962 · reddit · 2026-11-17

A Reddit user kicked off a discussion about running CVPR pre-submission reviews with Claude or ChatGPT: no matter how many times they fix the issues the AI reports, the verdict always comes back as "borderline" with limited novelty. The complaint taps into a shared experience among researchers — LLM-simulated reviewers tend to give template-like, unhelpfully hedged assessments.

Original post →

More from Fun

Fun channel →