Hugging Face's logo

wzhouad
/
Llama3-Instruct-8B-WPO-HB-v2

Text Generation
Transformers
Safetensors
llama
alignment-handbook
conversational
text-generation-inference
Inference Endpoints
Model card Files Community
1
Llama3-Instruct-8B-WPO-HB-v2 / README.md
wzhouad's picture
wzhouad
Update README.md
afab71f verified
preview code
|
raw
history blame
841 Bytes
metadata
library_name: transformers
tags: []

Description

Llama3-Instruct-8B model finetuned by hybrid WPO (GPT-4-turbo + on-policy sampling + Ultrafeedback). Details in WPO: Enhancing RLHF with Weighted Preference Optimization.

In comparison to the Llama3-Instruct-8B-WPO-HB model, it employs an enhanced preference data construction method:

  1. Uses the response with the minimum score as the rejected one.
  2. When multiple outputs have the same highest score, the one with the shortest length is selected.
  3. When multiple outputs have the same minimum score, the one with the smallest length difference from the chosen output is selected.

License

This model is licensed under the Zoom software license and is permitted for use only for noncommercial, educational, or academic research purposes.