Jev against Claude Haiku 5.5 on 262 real decisions Jev project

Jev against Claude Haiku 5.5 on 262 real decisions is a Benchmarks and evals project built with Jev. On 262 labeled cases from live production decision systems

On 262 labeled cases from live production decision systems, Jev got 204 right and Haiku 5.5 got 177.

Category: Benchmarks and evals. Built with Jev.

Author: Stanislav Sorokin

Official source

More Jev projects are on Jev projects and apps. Related: Laya projects and apps.

Similar projects in Benchmarks and evals

• Jev against Fable 5.1, 100 builders
• Jev against Llama and DeepSeek
• jev-align
• Jev and Laya on a medical diagnostic test
• jev-architect
• Jev as an observer of agent runs
• jev-as-judge
• jev-calibrate