Align-SLM: Textless Spoken Language Models with Reinforcement Learning from AI Feedback

TL;DR

Align-SLM introduces a reinforcement learning framework that improves the semantic coherence and relevance of textless spoken language models by leveraging preference optimization and AI feedback, achieving state-of-the-art results.

Contribution

The paper presents a novel reinforcement learning approach with preference optimization to enhance semantic understanding in speech-to-speech models.

Findings

Achieves state-of-the-art performance on ZeroSpeech 2021 benchmarks.

Improves semantic coherence in speech generation.

Outperforms previous SLM methods in multiple metrics.

Abstract

While textless Spoken Language Models (SLMs) have shown potential in end-to-end speech-to-speech modeling, they still lag behind text-based Large Language Models (LLMs) in terms of semantic coherence and relevance. This work introduces the Align-SLM framework, which leverages preference optimization inspired by Reinforcement Learning with AI Feedback (RLAIF) to enhance the semantic understanding of SLMs. Our approach generates multiple speech continuations from a given prompt and uses semantic metrics to create preference data for Direct Preference Optimization (DPO). We evaluate the framework using ZeroSpeech 2021 benchmarks for lexical and syntactic modeling, the spoken version of the StoryCloze dataset for semantic coherence, and other speech generation metrics, including the GPT4-o score and human evaluation. Experimental results show that our method achieves state-of-the-art performance for SLMs on most benchmarks, highlighting the importance of preference optimization to improve the semantics of SLMs.