StudentBench: One AI Tutor Matched Expert Human GRE Tutoring at 918x Lower Cost Across 2,383 Participants
arXiv·medium signal
arXiv 2609.28470 (23 Sep) collects over 175,000 student-AI messages and compares AI tutoring, human tutoring and no tutoring on Quantitative and Verbal GRE questions across 2,383 participants. AI tutoring was statistically equivalent to expert human tutoring (p=.015), the best AI tutor beat the human tutor in five of seven GRE domains on average, and one tutor matched human gains at USD 0.0052 against 4.81 per session. A second study used 2,028 expert pairwise ratings of LLM lesson plans and practice problems.