agent handbook
1.0.0
? Awesome AI Agents
I always believe in open source and love to share all the knowledge points that I think are valuable and interesting to Agent during my work and study, and regularly write them into blogs to discuss and learn with everyone and make progress together.
We are also very welcome to contribute PR to continuously improve this blog and make it a real Agent Handbook.
We strongly recommend that you read this speech by Mr. Ng to get started with Agent Workflow:
ORPO proposes a very innovative method: fuse the model alignment stage and the SFT stage together to improve the model training method.
In the SFT stage, the aligned data is directly added to the training, and the model alignment ability is realized in the SFT stage.
解决的问题: This paper aims to improve the ability to provide a method of creating high-quality instructions following data sets, thereby improving the ability to learn instructions in different methods.
In this paper, we generate a function to detect whether the Response content is correct, thereby improving data quality.
The method of this paper is not very innovative, but it tells us to a certain extent: the importance of data quality.
