doodod/Turn-Detector-Qwen3-0.6B

The doodod/Turn-Detector-Qwen3-0.6B is a 0.8 billion parameter Transformer-based language model, instruction fine-tuned from Qwen3-0.6B. Developed by doodod, this model is specifically designed for semantic-level turn detection in voice chat pipelines, predicting the probability of an end-of-turn token to prevent premature interruptions caused by VAD pauses. It excels in both Chinese and English dialogue scenarios, addressing the challenge of accurately determining user input completion.