While many look at the size of the model, prompting techniques are the second vector of development that will significantly increase the value of LLM. A recent paper on a new prompting technique has achieved results that are significantly better than previous ones, and the difference between GPT3.5 and GPT4 models is negligible https://arxiv.org/pdf/2311.04254.pdf https://cdn.discordapp.com/attachments/1173652086160892004/1177633120925778060/image.png?ex=6573374f&is=6560c24f&hm=d4ef6c54bc2c37b94c2f9c7fbe6c406945536b26fc579afcf3cbabb11bffa496&