Exam results (ordered by GPT 3.5 performance)
GPT 4 ■
Estimated percentile lower bound (among test takers)
GPT 4 (no vision)
GPT 3.5 ■
Figure 4. GPT performance on academic and professional exams. In each case, we simulate the conditions and scoring of the real exam. Exams are ordered from low
laaia»
J/Л » '»хД J
/Ámú вт ir . í ■ : " ijá
M|
/tfÂl /Им iS* ^WF &JW7-«*.- В/?í fas™ -S3 £
gWf SM 3 (y у*Ч|^Ч, r, * :&(*£' /MmB л л Г AI e 3a^ 'a Íwm9¿fiL 1 — •—j^E
ir ШШ1 дау mfi/Æ Krll