Prompting an LLM as if the paper is already published can produce more balanced critique
nptacek · x · 2026-07-22
The thread argues that models critique papers more evenly if you present the work as already published and ask for a general assessment first, rather than framing it as a submission and explicitly requesting criticism.
It also suggests two variants:
- ask the model to review the paper as a whole, without priming it to hunt for faults;
- or give two opposing takes to a third session and ask it which side is stronger.
The author adds that presenting the work as something you’d like critiqued outside of work can also produce a more balanced response.
More from Apps
- Revid auto-generates a video from a watched GitHub repo with no human edits — tibo_maker · 2026-07-22
- AI search indexes 78 websites and structures the first 40 podcast episodes — nptacek · 2026-07-22
- SciClaw_ai turns a prompt into a scientific diagram and PNG in seconds — Faheem_uh · 2026-07-22
- AI helps a writer recover decades of Medium content after a long index build — nptacek · 2026-07-22
- Google removes left-panel pinning for Gems, frustrating power users — bigripdadddy · 2026-07-22
- PenEcho turns a 20,000×20,000 canvas into a shared workspace for humans and AI — JafarNajafov · 2026-07-22