SICL-AT: Another way to adapt Auditory LLM to low-resource task
This addresses the challenge of adapting auditory LLMs to low-resource or unfamiliar tasks, which is incremental as it builds on existing in-context learning approaches.
The paper tackles the problem of adapting auditory large language models to low-resource tasks by proposing SICL-AT, a post-training method that enhances in-context learning using high-resource speech data, resulting in consistent outperformance over direct fine-tuning in low-resource scenarios.
Auditory Large Language Models (LLMs) have demonstrated strong performance across a wide range of speech and audio understanding tasks. Nevertheless, they often struggle when applied to low-resource or unfamiliar tasks. In case of labeled in-domain data is scarce or mismatched to the true test distribution, direct fine-tuning can be brittle. In-Context Learning (ICL) provides a training-free, inference-time solution by adapting auditory LLMs through conditioning on a few in-domain demonstrations. In this work, we first show that \emph{Vanilla ICL}, improves zero-shot performance across diverse speech and audio tasks for selected models which suggest this ICL adaptation capability can be generalized to multimodal setting. Building on this, we propose \textbf{Speech In-Context Learning Adaptation Training (SICL-AT)}, a post-training recipe utilizes only high resource speech data intending to strengthen model's in-context learning capability. The enhancement can generalize to audio understanding/reasoning task. Experiments indicate our proposed method consistently outperforms direct fine-tuning in low-resource scenario.