What it does
qwen-rlcd trains Qwen3.5-0.8B to make Jev-style choices, scores, and binary judgments with calibrated confidence.
Benchmarks & research
Screenshot unavailable. Open the experiment ↗
qwen-rlcd trains Qwen3.5-0.8B to make Jev-style choices, scores, and binary judgments with calibrated confidence.
You can use this to build an educational app that takes reading tests. Your developer gives the app a long story and asks several questions at once. The app picks the best answers from a list without writing new words.
Your developer can try the included performance test. The project needs a powerful computer graphics chip to run quickly. The outside videos and pictures used for testing have strict usage rules.