AI paper review before submission
Check your own manuscript for gaps between claims, methods and results. A report points back to the draft so you can decide which corrections matter.
AI paper review before submission is particularly useful when a small discrepancy could change the conclusion. A result moving from 60 to 66 is six percentage points higher. If the abstract reports something different, paper-review can flag the wording and point to the table where you can settle it.
Give the agent the text you intend to submit. It reads it first as a sceptical reader, then looks more closely at the contribution, clarity, results, testing and method. Alongside errors it can demonstrate, the report records questions that the paper leaves open and the evidence needed to answer them.
What it does
- Trace an abstract claim back to its supporting experiment.
- Check that training, tuning and evaluation data stay distinct in the account of the method.
- Compare a highlighted best score with the direction of the metric.
- Read one sentence per paragraph to find a missing step in the argument.
- Locate citations, captions or appendix references that need attention.
Install and use
- Install the skill in your agent. Its parallel review uses a code interpreter and the host’s subagents.
- Supply the full manuscript and appendices. Ask, “Where does my text promise more than my results show?”
- Work through the reported locations. Check whether each proposed correction would change a finding or clarify its description.
FAQ
Source
Made by EvoScientist. Original skill: paper-review, published on GitHub under the Apache-2.0 licence.
View on GitHub