Here is the code repository for the paper From Hypothesis to Premises: LLM-based Backward Logical Reasoning with Selective Symbolic Translation.

Before running the code, please make sure to insert your own api_key into the LLM_response_self function in utils.py.

Use the following command to run HBLR experiments by replacing model_name and dataset_name with your desired values:

python ./HBLR.py --model_name [your_model_name] --data_path ./data --dataset_name [your_dataset_name] --split dev

Use the following command to run Reasoning Verification by replacing model_name and dataset_name with your desired values:

python ./verifier.py --model_name [your_model_name] --data_path ./data --dataset_name [your_dataset_name] --split dev

The dataset_name argument supports the following options: FOLIO, ProntoQA, ProofWriter, LogicalDeduction, and AR-LSAT. Replace [your_dataset_name] in the command with one of these dataset names to run the corresponding HBLR experiment.