GAIN: Multiplicative Modulation for Domain Adaptation
This addresses the issue of forgetting in domain adaptation for LLM users, offering a more stable and efficient approach, though it is incremental as it builds on existing modulation techniques.
The paper tackles the problem of catastrophic forgetting in domain adaptation for large language models by proposing GAIN, a multiplicative modulation method that re-emphasizes existing features, resulting in improved performance on previously trained domains by 7-13% in perplexity and minimal degradation in downstream tasks compared to LoRA.
Adapting LLMs to new domains causes forgetting because standard methods (full fine-tuning, LoRA) inject new directions into the weight space. We propose GAIN, which re-emphasizes existing features through multiplicative modulation W_new = S * W. The learned diagonal matrix S is applied to the attention output projection and optionally the FFN. The principle mirrors gain modulation in neuroscience, where neurons adapt to context by scaling response strength while preserving selectivity. We evaluate GAIN on five models from four families (774M to 70B), adapting sequentially across eight domains. GAIN-FFN matches LoRA's in-domain adaptation, but their effects on previously trained domains are opposite: GAIN-FFN improves them by 7-13% (validation PPL), while LoRA degrades them by 18-36%. Downstream accuracy confirms the pattern: for example, after seven sequential adaptations on Qwen2.5, GAIN-FFN degrades BoolQ by only 0.8% while LoRA damages it by 14.9%. GAIN adds 46K-230K parameters per model and can be absorbed into the pretrained weights for zero inference cost.