LoRAMoE: Alleviate World Knowledge Forgetting in Large Language Models via MoE-Style Plugin
This addresses a key issue for developers and users of large language models by preventing degradation of stored knowledge during fine-tuning, though it is incremental as it builds on existing adapter and MoE techniques.
The paper tackles the problem of world knowledge forgetting in large language models during supervised fine-tuning with increased instruction data, and proposes LoRAMoE, a framework that uses low-rank adapters and a router network to maintain world knowledge while improving downstream task performance, with experimental results showing significant improvements.
Supervised fine-tuning (SFT) is a crucial step for large language models (LLMs), enabling them to align with human instructions and enhance their capabilities in downstream tasks. Increasing instruction data substantially is a direct solution to align the model with a broader range of downstream tasks or notably improve its performance on a specific task. However, we find that large-scale increases in instruction data can damage the world knowledge previously stored in LLMs. To address this challenge, we propose LoRAMoE, a novelty framework that introduces several low-rank adapters (LoRA) and integrates them by using a router network, like a plugin version of Mixture of Experts (MoE). It freezes the backbone model and forces a portion of LoRAs to focus on leveraging world knowledge to solve downstream tasks, to alleviate world knowledge-edge forgetting. Experimental results show that, as the instruction data increases, LoRAMoE can significantly improve the ability to process downstream tasks, while maintaining the world knowledge stored in the LLM.