<ul><li>Current parameter-efficient fine-tuning methods for adapting pre-trained language models to downstream tasks are susceptible to interference from noisy data.</li><li>Proposed a noise-robust adaptation method via asymmetric LoRA poisoning experts (LoPE) framework that enhances model robustness with generated noisy data.</li><li>LoPE strategically integrates a dedicated poisoning expert in an asymmetric LoRA configuration to enhance noise discrimination and processing ability.</li><li>Extensive experiments demonstrate that LoPE achieves strong performance and robustness purely through low-cost noise injection, eliminating the need for data cleaning.</li></ul>

Noise-Robustness Through Noise: Asymmetric LoRA Adaption with Poisoning Expert

Discover more