What it does
Its paired Chinese and English prompts make language differences visible alongside GridWorld and chess experiments.
Benchmarks & research
A Chinese-language experiment suite for Jev's context limits, bilingual judgments, batching and game decisions.
Screenshot unavailable. Open the experiment ↗
Its paired Chinese and English prompts make language differences visible alongside GridWorld and chess experiments.
Start by collecting pairs of matching English and Chinese sentences, along with simple multiple-choice questions about them. You can have a developer run these test scripts to see how well the model reads both languages or chooses moves in simple games.
Your developer will need a connection key from a paid OpenRouter account. Running these scripts sends real requests across the internet, so each test run adds to your bill. These tests only compare specific written examples and do not guarantee overall accuracy.