AI tutoring matches human GRE learning gains at 918x lower cost, 2,383-person study finds

handshake-ai-research · hf · 2026-09-24

handshake-ai introduces StudentBench, an evaluation suite and public platform (175,000+ student-AI messages) studying whether LLMs deliver human-equivalent tutoring gains.

Key findings:

Across 2,383 participants receiving AI tutoring, human tutoring, or none, AI tutoring was statistically equivalent to expert human tutoring on GRE learning gains (p=.015); in five of seven GRE domains the best AI tutor beat the human on average.
Cost gap: the best AI tutor achieved equivalent gains (p=.044) at 918x lower cost ($0.0052 vs $4.81 per percentage point gained).
A second study had expert tutors run 2,028 pairwise rubric evaluations of LLM lesson plans and practice problems, separating AI tutors on lesson planning, problem creation, conversational pedagogy, cost, and engagement.
In Quantitative GRE sessions, faster AI replies correlated with more student messages, more correct practice, and larger gains (all p<.002).

The platform is freely available at studentbench.org.

Original post →

AI tutoring matches human GRE learning gains at 918x lower cost, 2,383-person study finds

More from Apps