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.
Researchers introduce RetroThinker, a multi-stage post-training framework that equips the Moshi speech LLM to verify and forward-correct chain-of-thought steps during streaming inference. It combines supervised fine-tuning on curated retrospective thinking data with length-based direct preference optimization (DPO). On GSM8K it achieves an 11% absolute accuracy gain over non-retrospective baselines at comparable latency.
- Targets the accuracy-latency trade-off in real-time spoken interaction.
- Combines SFT with length-based DPO for retrospective reasoning.
- Enables self-verification and forward-correction of CoT steps during inference.
- 11% absolute GSM8K accuracy gain at comparable latency.
- Products
- RetroThinker
- AI models
- Moshi
From the original article
1 min read at arxiv.orgSpeech large language models (SpeechLLMs) offer reduced latency and retain paralinguistic nuances that are typically lost in cascaded automatic speech recognition (ASR) and text-based LM architectures. However, they continue to lag behind text-only LLMs on complex reasoning tasks, while real-time spoken interaction imposes strict latency constraints.
By Yi-Jen Shih · arxiv.org
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.