Main documentation instruction image needs a description

[Kallinteris-Andreas]

From what I can tell, the image on https://huggingface.co/docs/trl/index is based on the instructGPT paper.

Adding a figure description (like LaTeX papers do) stating that this is a sample post-training technique, and it is based on the instructGPT paper and modern post-training methods vary, should be good enough

or remove it

btw paper title of instructGPT is: "Training language models to follow instructions with human feedback"