RetroThinker: Enabling Retrospective Thinking in Speech LLMs
RetroThinker is a post-training framework letting the Moshi speech LLM self-correct reasoning mid-stream, adding 11% GSM8K accuracy at similar latency.
1 stories, newest first · one card per story
RetroThinker is a post-training framework letting the Moshi speech LLM self-correct reasoning mid-stream, adding 11% GSM8K accuracy at similar latency.
Today 1.6K readers · 2.5K views · 780 reads · all time 189K views