Designed for Inspirit AI
Rubric-based writing review, built for a research program's own deliverables
Inspirit AI's AI + X research program pairs high-school students with a research mentor; students write a project proposal, a milestone report, and a final paper. GradPilot built one rubric for each of those three documents, from what science-fair judges, student research journals, and university course-project guides publish about what they look for.
What your students get
Research Proposal
Reviews your project proposal the way research mentors and science-fair readers read one. Paste the whole proposal, template headings included.
What it looks for
- The Question. Whether the proposal states one research question that names what will be studied and what will be measured or compared, and that a result could answer.
- The Data. Whether the proposal names its data, where the data came from, how much there is, what one record looks like, and what the data covers.
- The Method and What's Yours. Whether the proposal names a method, matches it to the question, names the implementation it starts from, and separates the writer's own work from what the program, mentor, or existing code provides.
- The Evaluation Plan. Whether the proposal names a measure that fits the task, a number to compare it against, a plot, and a way of looking at individual results.
Project Milestone
Reviews your milestone report the way course project readers read one. Paste the whole document, template headings included.
What it looks for
- The Problem, Restated. Whether the milestone restates the problem as a question with a stated input and output, gives a sourced reason it matters, and uses the same target the data and results sections use.
- The Data Pipeline. Whether the milestone names the data and its origin, the preprocessing, the split with its rule and its unit, and the shape of what the model sees.
- Preliminary Numbers. Whether the milestone reports at least one result with the metric defined, the model and split it came from, and something to compare it against.
- Next Steps and Risks. Whether the remaining work is named step by step, tied to what the results showed, ordered or timed, and guarded by a fallback.
Research Paper
Reviews your research paper the way student journals, science fairs, and symposium judges read one. Paste the whole paper, headings included, at any stage of drafting.
What it looks for
- Abstract and Question. Whether the abstract states the problem, the method, a result with a number, and a conclusion, and whether the paper states one research question whose target the results answer.
- Methods Another Student Could Repeat. Whether the methods give the data's origin and size, the preprocessing and split with its rule, the model with how it was trained, the settings and how they were chosen, and the code and tools it used.
- Results With a Comparison. Whether every result has a named measure, a model and split it came from, a comparison on the same measure, and a figure or table the text actually reads.
- Claims Inside the Evidence. Whether each conclusion points at a result in the paper, each generalizing claim carries a condition, limitations name a property of the data or method, and failures are explained.
- A free Quick Review each day. Two per day. A Quick Review returns critiques — what is not working, tied to the student's own sentences.
- A Full Review. Adds a score for every section of the rubric and specific suggestions for fixing what it flags.
- A review a student can share. Any review can be shared by link with a mentor, an advisor, or a teacher.
- Mentor comments on a shared review. Whoever opens the link can respond to the same critiques — agree or disagree, and leave a comment — under a display name, with no account.
What your program gets
- One rubric per deliverable. Each document your students hand in gets its own rubric, mirroring the criteria those students are actually judged by.
- The same standard every time. Every student gets the same reading of the same document, from every mentor and every review.
- The standard comes first. Each section says what it is reading for before it says anything about the draft, so a student can see the standard before the feedback.
- It never writes the student's text. The review returns critiques and suggestions on the writing that was submitted. It does not draft or rewrite any part of the document.
- Venue rules on AI are quoted, not enforced. Where a fair or a journal publishes a rule about AI use or about what counts as your own work, the rubric quotes it as information. It never asks for or rewards a statement about AI use, never treats a venue's rules as a reason a document cannot be submitted, and never suggests working around a rule a student has agreed to.
- Scores are revision signals. They point at what to revise next. They are not grades, and they do not predict what a mentor, a journal, or a fair will decide.
How we build one for a program
- Read the program's own documents. We start from the deliverable templates a program hands its students, and from the published criteria of the readers those documents go to.
- Build one rubric per deliverable. Each rubric covers one document, and cites inside it the sources its sections came from.
- Audit before release. A reviewer who did not build the rubric audits it against those sources before it goes live.
- Run it end to end. We run every rubric on sample documents at different stages of drafting and read what it actually flags.
What we read
These rubrics mirror the criteria these readers publish. Listing a source means we read what it publishes — none of the organizations below reviewed or approved these rubrics.
Science fairs and competitions
- Society for Science — Regeneron ISEF Grand Award Judging Criteria
- Society for Science — Regeneron ISEF Rules for All Projects
- Society for Science — How to Write an ISEF Abstract
- Society for Science — ISEF Rules Educator Guide (2025–2026)
- Society for Science — ISEF Form 1A Student Checklist and Research Plan Instructions (2025–2026)
- Regeneron Science Talent Search — Research Report Guidelines (2027)
- JSHS — Core Rules of Competition 2026–27
- North Carolina Science and Engineering Fair — 2027 Judging Handbook
Student research journals
University course-project guides
- Stanford CS230 — Project (proposal, milestone, and final report guidelines)
- Stanford CS229 — Final Report Guidelines
- Stanford CS229 — Project Guidelines (Spring 2016)
- Stanford CS224n — Project Proposal Instructions
- Stanford CS224n — Project Milestone Instructions (Spring 2024)
- Stanford CS224R — Custom Project Guidelines
- Cornell CS 5757 — Project Milestone Guidelines (Spring 2026)
- University of New Hampshire Hamel Center — Writing an Effective Research Proposal
- George Mason University Writing Center — How to Write a Research Question
Research on machine-learning practice
Bring this to your program
We build rubrics like these for the documents a program already assigns — proposals, milestone reports, final papers. A short call is enough to see what your deliverables would need.