After the first paper that has demonstrated that language models are great few-shot learners.
Language Models are Few-Shot Learners
Prompting refers to an idea of using a customized function to help big language models to become effective on down-stream tasks.
Consider a pre-trained language model now is working on a prediction task that produces a class-wise prediction likelihood of $p(y|x_i)$, we can effective have a prompting function $f_{prompt}$ so that now the prediction task becomes $p(y|f_{prompt}(x_i))$.
There are a few great papers in 2020 and 2021, describing in details how prompting would work to help boost the performance of large language models such as GPT3:
GitHub - ucinlp/autoprompt: AutoPrompt: Automatic Prompt Construction for Masked Language Models.
Differentiable Prompt Makes Pre-trained Language Models Better...
Most prompting methods are based on texts. Even the very recent paper related to this project describes a method that works on vision-language models such as CLIP; the prompt is still a text prompt, integrated into the text input part of the CLIP model.
Conditional Prompt Learning for Vision-Language Models
There are currently a variety of augmentation methods in computer vision that focus on providing a wide range of augmentations. This allows the training stage to become exposed to a vast array of samples, which in turn results in the final trained model having better generability on unseen samples.
Here is a collection of data augmentation methods