RankMean: Module-Level Importance Score for Merging Fine-tuned LLM Models.
Gabriel PerinXuxi ChenShusen LiuBhavya KailkhuraZhangyang WangBrian GallagherPublished in: ACL (Findings) (2024)
Keyphrases
- fine tuned
- artificial intelligence
- probabilistic model
- neural network
- modeling framework
- mathematical models
- statistical models
- prior knowledge
- pairwise
- case study
- markov random field
- model selection
- generative model
- decision making
- information systems
- machine learning
- fine tuning
- abstraction levels
- accurate models
- data sets