of t is too small, such as only a

[rifattryo.ut11]

As of , -t already supports up to t. Here it represents one thousand. That is, thousands of t. How to write prompts for large language models elegantly and economically has become a craft. As of , the T model and its context length limit in , Giving instructions to large language models is a bit like inputting instructions into early computers. Before the invention of keyboards, you need to feed instructions to the computer bit by bit on a punched paper tape h. This is how Bill Gates and Paul Alan operated computers when they first started to collaborate on programming.



Early M punched paper tape rows and columns iran telephone number can input roughly byte instructions. - can input byte instructions. So the question is, which language can express the most meaning with the least t? Chinese? The treasures left to us by our ancestors have always been concise and to the point. Does this mean that Chinese can be used as the command prompt language of htT? If we go a step further, is there any language in the world that is more "concise" than Chinese? Ranking of language conciseness From Chinese to English to Spanish to Japanese, there are many studies that explore how to determine the conciseness of various languages ​​in the world from different perspectives.



Here we cite two studies for your reference. The length of the same text after translation Someone evaluated the different language translations of Google's privacy policy on the public Internet. The following are examples of languages ​​ranked by the total number of characters. Traditional Chinese characters. Simplified Chinese characters. Japanese characters. English characters. Spanish characters. French characters. Vietnamese characters. Hindi characters. This can quantify the information density of each language. Traditional and Simplified Chinese are indeed very concise languages! But htT doesn't think so. What is the most effective language? This table shows the translation of the Google privacy policy snippet in different languages.