sDPO: Don't Use Your Data All at Once

Published on Mar 28
· Featured in Daily Papers on Mar 29


As development of large language models (LLM) progresses, aligning them with human preferences has become increasingly important. We propose stepwise DPO (sDPO), an extension of the recently popularized direct preference optimization (DPO) for alignment tuning. This approach involves dividing the available preference datasets and utilizing them in a stepwise manner, rather than employing it all at once. We demonstrate that this method facilitates the use of more precisely aligned reference models within the DPO training framework. Furthermore, sDPO trains the final model to be more performant, even outperforming other popular LLMs with more parameters.


why it works?

This is an automated message from the Librarian Bot. I found the following papers similar to this paper.

The following papers were recommended by the Semantic Scholar API

Please give a thumbs up to this comment if you found it helpful!

If you want recommendations for any Paper on Hugging Face checkout this Space

You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend

It's not clear to me from the Ablation whether this is just a function of learning rate cycling - like has been well explored in the CV literature.

Sign up or log in to comment

Models citing this paper 5

Browse 5 models citing this paper

Datasets citing this paper 0

No dataset linking this paper

Cite in a dataset to link it from this page.

Spaces citing this paper 29

Collections including this paper 18