Kyutai Releases Voice of Reason: A Speech-Native Model that Solves Spoken Math with Reinforcement Learning
MarkTechPost
Read Full Article at MarkTechPost →Ad Slot — In-Article (728x90)
Kyutai has released Voice of Reason, 2 open-weight speech-to-speech models built on GLM-4-Voice-9B. Supervised fine-tuning and reinforcement learning lift spoken GSM8K accuracy from 27. 3% to 77. 1%. There is no transcription step and no text LLM in the loop.
Both checkpoints are on Hugging Face and run on a single H100. The post Kyutai Releases Voice of Reason: A Speech-Native Model that Solves Spoken Math with Reinforcement Learning appeared first on MarkTechPost.
This is a summary. For the full story, read the original article at MarkTechPost.
Original source: MarkTechPost