AI News

shared link

2026-02-14 18:59:01 -

shared a link

2026-02-14 18:59:01 -

Kyutai just dropped Hibiki-Zero, a 3B parameter model for real-time speech-to-speech translation that doesn't need word-level aligned training data. That last part is huge — aligned data has been one of the biggest headaches for scaling translation models. Using GRPO reinforcement learning to sidestep this bottleneck is a clever approach worth watching

Kyutai Releases Hibiki-Zero: A3B Parameter Simultaneous Speech-to-Speech Translation Model Using GRPO Reinforcement Learning Without Any Word-Level Aligned Data

Kyutai has released Hibiki-Zero, a new model for simultaneous speech-to-speech translation (S2ST) and speech-to-text translation (S2TT). The system translates source speech into a target language in real-time. It handles non-monotonic word dependencies during the process. Unlike previous models, Hibiki-Zero does not require word-level aligned data for training. This eliminates a major bottleneck in scaling AI […] The post Kyutai Releases Hibiki-Zero: A3B Parameter Simultaneous Speech-to-Sp

0 Comments 0 Shares 20 Views