metadata
base_model: unsloth/meta-llama-3.1-8b-instruct-bnb-4bit
library_name: peft
Control Summaries LLaMA (LoRA adapter)
This is a LoRA adapter for controllable summarization trained for the paper Learning to Control Summaries with Score Ranking.
Model Details
- Model type: LoRA adapter for a causal language model (summarization control)
- Language(s) (NLP): English
- Finetuned from model:
unsloth/meta-llama-3.1-8b-instruct-bnb-4bit
Model Sources
- Repository: https://github.com/Hyfred/Control-Summaries-with-Ranking
- Paper: Learning to Control Summaries with Score Ranking
Uses
Direct Use
Use this adapter with the base model for controllable summarization (more complete / more concise / balanced prompts).
How to Get Started
Please refer to the code repository for usage and evaluation scripts: https://github.com/Hyfred/Control-Summaries-with-Ranking
Bias, Risks, and Limitations
The model inherits limitations and potential biases from the base model and training data. Summaries can be incomplete or contain errors. Review outputs before use in critical settings.
- PEFT 0.14.0