JEV-as-a-Judge Nears GPT-6 Accuracy at a Fraction of Cost
JEV-as-a-Judge, a probability-only decision model, ranks 8th among 15 judges while matching GPT-6's accuracy within 1.7 points at 277x lower cost. A CMU paper shows a confidence-routing cascade that beats GPT-6 accuracy at 41% of its cost.
2026-10-11 ~ 2026-10-11 · 2 related posts
- JEV judge is only 1.7 points behind GPT-6 yet costs 277x less — kalyan_kpl · 2026-10-11
- JEV-as-a-Judge: confidence-routed cascade beats GPT-6 accuracy at 41% of the cost — kalyan_kpl · 2026-10-11