Eval-G1

uses registered agent steps and tool calls, together with feedback from subject matter experts, to create small evaluation models

“Eval-G1 gebruikt de geregistreerde stappen en toolaanroepen van agents, samen met feedback van vakexperts, om kleine evaluatiemodellen te maken.”

promises up to 98 percent lower evaluation costs than a commercial language model and evaluations 2.5 times faster than traditional language model reviewers

“Eval-G1 tot 98 procent lagere evaluatiekosten en 2,5 keer snellere evaluaties belooft”