# What is Post-training?

It is the improvement process performed for a model that has completed its basic training to perform better on specific tasks.

## Overview
After pre-training, where the model gains its general capabilities, this is an additional stage performed for the model to adopt a specific style or specialize in a particular subject. During this process, the model is trained with higher-quality data and refined to remove its errors.

*Analogy: It is like a university graduate gaining sectoral experience through an internship before starting their professional career.*

## How it works
The model is retrained using methods such as RLHF or specific datasets. This ensures it provides safer, more accurate, and more useful responses.

## Where it is used
It is used in the preparation stage of large language models before they are presented to the end user.

## Commonly confused with
It is confused with pre-training; while pre-training lays the foundations of the model, post-training shapes the behavior of the model.

## Frequently asked questions
**Why is post-training necessary?**
It is necessary to fine-tune the model's behaviors so that it provides safer and more user-friendly responses.


## Related terms
- [Fine-tuning](/en/dictionary/fine-tuning/)
- [RLHF](/en/dictionary/rlhf/)
- [Foundation Model](/en/dictionary/foundation-model/)

## Related tools
- [Miles](/en/discover/miles/)

---
Source: TreScout Dictionary · https://trescout.com/en/dictionary/post-training/