What it does
The fixed game datasets cover support-triage routing, attack-type labeling, and a large banking-intents menu, with fairness rules pinned in a written guide so neither contender gets a softer prompt.
Benchmarks & research
An open battleground that feeds Jev and Laya byte-identical rounds — same wording, same option order — and publishes the report card.
Only you can see your notes.
Screenshot unavailable. Open the experiment ↗
The fixed game datasets cover support-triage routing, attack-type labeling, and a large banking-intents menu, with fairness rules pinned in a written guide so neither contender gets a softer prompt.
A developer can use this code to test two AI programs side by side. The app sends the exact same customer ticket to both. It compares an outside AI service called Jev against a local program called Laya. Finally, it builds a visual report card.
A developer can add new test questions to the games file. The app needs an access key to connect to Jev. Your computer also needs extra hard drive space to download Laya. The actual testing speed will depend on your local hardware.