wenhuchen/TableCoT

The code and data used for EACL2023 Paper: "Large Language Models are few(1)-shot Table Reasoners"

TableCoT

The code and data used for EACL-2023 Paper Large Language Models are few(1)-shot Table Reasoners

Preliminary

First, you need to specify your OPENAI_API_KEY, please find it in your account in https://openai.com/api/.

export OPENAI_KEY=[YOUR_KEY]

For WikiTableQuestions

python prompt.py --start 0 --end 500

This will call Chain of Thoughts prompting to solve the 0-500 example in the test set of WikiTableQA. The output will be saved to output/response_..._s0_e500.json.

You can further call this following to extract the answers from the predictions.

cd outputs/
python postprocess_answer.py --inputs response_..._s0_e500.json

Finally, call this following to compute the final EM score.

python compute_scores.py --inputs response_..._s0_e500.json.processed

For TabFact

python prompt.py --start 0 --end 500

This will call Chain of Thoughts prompting to solve the 0-500 example in the test set of WikiTableQA. The output will be saved to output/response_..._s0_e500.json. This will directly output the accuracy after it finishes.