metadata
library_name: transformers
tags:
- trl
- dpo
- first-order-logic
datasets:
- yale-nlp/FOLIO
language:
- en
base_model:
- meta-llama/Llama-3.2-3B
Model Card for Model ID
Aligned version of meta-llama/Llama-3.2-3B for Logical Reasoning
Model Description
This is an aligned model using DPO in order to improve the base model's performance in formal reasoning in first-order logic.
- Developed by: [Grupo de Ingeniería Lingüística]
- Language(s) (NLP): [English]
- License: [Whichever one Llama 3 uses]
- Finetuned from model [meta-llama/Llama-3.2-3B]:
Model Sources [optional]
- Repository: [Github]
- Paper: Into The Limits of Logic: Alignment Methods for Formal Logic Reasoning
Evaluation
Citation [optional]
BibTeX:
[More Information Needed]