Create your account

StudentBench is a free AI tutoring service for the GRE, run by Handshake AI Research. We ask for your first and last name, email, and age to create your account. Inside the app you are shown only as an anonymous ID number; your name is stored separately for the research team.

Plan to finish in one sitting (about 2–2.5 hours) — the timer keeps running once you start, though there are short breaks.

First name Last name Email Age Password

Informed Consent Form

StudentBench: AI Tutoring Research Study for GRE and SAT Preparation
For adult participants ages 18 and older (GRE participants, and SAT participants age 18 or older)

Study Title	StudentBench: Benchmarking LLM Tutoring Effectiveness on Standardized Tests (SAT & GRE)
Principal Investigator	Curtis Northcutt, PhD, Handshake AI (Stryder Corp. d/b/a Handshake)
Sponsor	Handshake AI (internal research)
Protocol #	26-206

Key information about this study

The following is a short summary to help you decide whether to take part. More detail is provided in the rest of this form.

You are being invited to take part in a research study because you are preparing for the GRE and/or SATs. Taking part is completely voluntary.

The purpose is to measure how much an AI tutor improves GRE and/or SAT practice scores and to compare different AI tutoring systems.

Participation takes about 2–2.5 hours in a single online session. Students will complete a pre-test, approximately 1-hour tutoring segment, and a post-test. You will be randomly assigned to an AI tutoring model.

Risks

Participation involves minimal risk. The main potential discomforts are mild frustration during the test and the time required (2–2.5 hours). The tests do not count toward any grade or academic record. Your assigned tutor may occasionally make mistakes; this is part of what we are studying. The alternative is to not take part in this research study.

Benefits

You will receive a personalized performance report. You will also receive free GRE or SAT tutoring on your weak areas. There are no guaranteed academic benefits, but many participants find the practice session genuinely helpful.

Purpose of the Study

You are being invited to participate in a research study about how well AI systems can tutor people studying for the GRE or SAT. The goal of this study is to measure whether, and by how much, an AI tutor improves a student's test score, and to compare different AI tutoring systems. Results will be used to build a public leaderboard that helps AI labs develop better educational tools.

What You Will Be Asked to Do

If you agree to participate, you will:

Complete pre-test on GRE Quantitative or SAT topics (about 1 hour).
Take part in an approximately 60-minute tutoring segment. You will be randomly assigned to one of several AI systems. You do not choose your model. The tutoring focuses on the topics you missed on the pre-test.
Complete post-test on equivalent GRE or SAT topics.
Optionally provide brief feedback on your experience.

Total time commitment: approximately 2–2.5 hours. Everything takes place online.

What We Will Measure

The primary outcome is your improvement from pre-test to post-test. We also collect your name, email, age, and grade level, along with data about your tutoring session (chat logs or session transcripts, where applicable) and behavioral signals including time spent per question, response latency, and number of tutor turns.

Confidentiality

You provide your name, email, age, and grade level, which are collected solely for the purposes of this study.
These identifiers are kept confidential, stored separately from your study answers, and replaced with a random study ID that only the research team can link.
Your chat logs and test scores are stored on a secured, access-controlled study server (studentbench.org) and in Handshake's secure enterprise data systems.
Your data will be used only for this research study and related academic publications.
Aggregate or de-identified findings may be published in academic papers or public benchmarks. No individual participant will be identifiable in any publication.
Identifiers might be removed from your identifiable private information. After such removal, the information could be used for future research studies or distributed to another investigator for future research studies without your additional informed consent.

Your rights and choices

Your participation is entirely voluntary. You can withdraw at any time, for any reason, with no consequences to your work on Handshake. Once data has been de-identified and incorporated into analysis, it may no longer be technically possible to isolate and remove it.

On publication, de-identified study materials (such as aggregate results, AI tutor rankings, the practice-question bank, and analysis code) may be released publicly as part of open-science practices. Any chat transcripts or scores would be shared only in de-identified and/or aggregated form.

Your information is retained for 6 years following publication, then securely deleted. Before any results are shared publicly, all identifying information is removed; you are never identified by name.

Voluntary Participation

Participation is completely voluntary. You may stop at any time without penalty. Withdrawing will not affect your relationship with Handshake or any other institution.

Questions?

If you have questions about this study, contact Inaara Hasmani at [email protected]. If you have any questions or complaints, you may contact a person not on the research team at the Biomedical Research Alliance of New York Institutional Review Board at (516) 318-6877 or at www.branyirb.com/concerns-about-research.

By checking the consent box below this form, I confirm that I am at least 18 years of age, that I have read and understood this form, and that I voluntarily agree to take part in this study.

Version: A, B · Protocol #: 26-206 · IRB APPROVED 07/06/2026

Already have an account? Log in.