Prompt-based methods involve using small, task-specific input modifications (prompts) to guide a model's behavior, particularly in scenarios like continual learning. They enable models to adapt to new tasks and acquire knowledge while retaining previously learned information, often by conditioning the model's output or internal representations.
Prompt-based methods help AI models learn new information without forgetting old knowledge, especially in scenarios where models need to adapt to many tasks over time. They work by adding small, task-specific cues (prompts) to the input, guiding the model's behavior for each task. This approach improves model adaptability and efficiency, particularly in continual learning.
Soft prompts, Prefix tuning, Prompt tuning, Prompt engineering, Adapter-based methods
Was this definition helpful?