AnySimLite: A Lightweight Few-Shot Similarity Encoder for On-Device Speech-Adjacent Classification
It addresses the memory footprint challenge of deploying multiple specialized models on edge devices by enabling a single lightweight architecture to handle various speech-adjacent classification tasks.
AnySimLite is a lightweight similarity encoder that achieves state-of-the-art or competitive performance on multiple speech-adjacent classification tasks in few-shot settings, with a performance drop below 7% while using less than 1/250th the model size of the SOTA qLLaMA_LoRA-7B baseline.
To minimize privacy concerns and inference latency on edge devices like smartphones, lightweight on-device models remain important for end-user applications. Many of these applications involve natural language classification, but deploying multiple specialized models creates a memory footprint challenge. We investigate: Can a single lightweight architecture solve multiple Speech-Adjacent (SA) classification tasks through reduction to a nuanced text similarity formulation? We propose AnySimLite, a lightweight similarity encoder that combines word-level and character-level channels. Together with a dataset transformation strategy, we evaluate AnySimLite across multiple SA classification tasks and show that it consistently achieves state-of-the-art (SOTA) or SOTA-competitive performance in few-shot settings while maintaining a low memory footprint. Even in the worst case, the performance drop remains below 7% while using $<\frac{1}{250}^{\mathrm{th}}$ of the model size of the SOTA qLLaMA_LoRA-7B baseline.