What are Adapters

Adapters are lightweight neural network modules inserted into pre-trained models to enable task-specific adaptation without modifying the original model parameters. These parameter-efficient fine-tuning techniques add small trainable layers while keeping the base model frozen, allowing rapid customization for new domains or tasks. Common adapter architectures include bottleneck adapters with down-projection and up-projection layers, Low-Rank Adaptation (LoRA) that decomposes weight updates into low-rank matrices, and prefix tuning that prepends learnable tokens to input sequences. Adapters typically add less than 5% additional parameters while achieving performance comparable to full fine-tuning. Benefits include reduced computational costs, faster training, prevention of catastrophic forgetting, and support for multi-task learning where different adapters handle different capabilities. For AI agents, adapters enable efficient personalization, domain adaptation, and skill acquisition without expensive retraining, making systems more modular and cost-effective.