Skip to content

Rate, Explain and Cite (REC): Enhanced Explanation and Attribution in Automatic Evaluation by Large Language Models. Paper: https://arxiv.org/pdf/2411.02448

Notifications You must be signed in to change notification settings

adelaidehsu/REC

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

14 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Rate, Explain and Cite (REC): Enhanced Explanation and Attribution in Automatic Evaluation by Large Language Models

Rec-Data

Our curated citations and explanation fine-tuning data is provided as REC_data.zip, which contains 24,081 (prompt, completion) pairs. Note that the full training data of the REC models additionally included HelpSteer2, Skywork, OffsetBias, and Code Preference. The additional data can be downloaded at their own websites.

Model training

If you wish to start the training from scratch -- the SFT trainer scripts for both REC-12B and REC-70B can be found at the training folder. Please put the unzipped REC_data at the correct path as you specify in the file, and modify the output_dir to be where you wish the trained model to be saved.

About

Rate, Explain and Cite (REC): Enhanced Explanation and Attribution in Automatic Evaluation by Large Language Models. Paper: https://arxiv.org/pdf/2411.02448

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages