Seal: Advancing Speech Language Models to be Few-Shot Learners
This work addresses the challenge of few-shot learning in multi-modal settings for speech and language processing, representing an incremental advancement.
The paper tackles the problem of extending few-shot learning capabilities to speech-language tasks by introducing the Seal model, which uses a novel alignment method with KL divergence loss to bridge frozen speech and language components, achieving robust performance on two speech understanding tasks.
Existing auto-regressive language models have demonstrated a remarkable capability to perform a new task with just a few examples in prompt, without requiring any additional training. In order to extend this capability to a multi-modal setting (i.e. speech and language), this paper introduces the Seal model, an abbreviation for speech language model. It incorporates a novel alignment method, in which Kullback-Leibler divergence loss is performed to train a projector that bridges a frozen speech encoder with a frozen language model decoder. The resulting Seal model exhibits robust performance as a few-shot learner on two speech understanding tasks. Additionally, consistency experiments are conducted to validate its robustness on different pre-trained language models.