The importance of prompt design: Examples first, instructions behind.
Haebom
One sentence to remember when giving AI commands through prompts. Put example data in front, instructions later. (EFIL: examples first, instructions later) The reason is simple. Due to the nature of language models, they tend to be biased towards information they have recently received. To use this bias to our advantage, it is better to put instructions at the end of the prompt.
Location of examples and instructions
If you put examples before the prompt, the language model won't rely too much on those examples. If you put instructions after, the language model will be better at following those instructions.
Simplicity and effectiveness of structure
You can achieve effective results while keeping the structure simple. This structure helps reduce repetition and off-topic talk. In other words, the user's input style can have a big impact on the output of the language model. For example, if the input is complex and ambiguous, the model's output can be complex and ambiguous accordingly. Conversely, clear and specific input can help the model produce more accurate output.
User input and language model response
User input can come in a wide variety of forms and styles. It can include questions, instructions, or simple keywords. Language models need to respond appropriately to this variety of input. Placing examples in the front, input in the middle, and instructions in the back can help the model respond more accurately and consistently.
Subscribe to 'haebom'
📚 Welcome to Haebom's archives. --- I post articles related to IT 💻, economy 💰, and humanities 🎭. If you are curious about my thoughts, perspectives or interests, please subscribe. Would you like to be notified when new articles are posted? 🔔 Yes, that means subscribe. haebom@kakao.com
