Disentangling Length from Quality in Direct Preference Optimization.
Ryan ParkRafael RafailovStefano ErmonChelsea FinnPublished in: ACL (Findings) (2024)
Keyphrases
- higher quality
- high quality
- optimal design
- constrained optimization
- optimization process
- global optimization
- discrete optimization
- quality assessment
- optimization method
- user preferences
- data sets
- real time
- quality measures
- optimization algorithm
- database systems
- website
- low quality
- artificial intelligence
- neural network
- fixed length