Here is an example to run qasper dataset with gpt-3.5-turbo

### Running

First, change the key in `gpt.py` and setup the commands in `scripts/gpt3.5.sh`

Then, `bash scripts gpt3.5.sh`

You can see the results show up after running is done

### Evaluation

To evaluate the results, find the result in `result/scrolls_qasper/gpt-3.5-turbo-0125/12000.json`, and go to `analysis/new_bench.py`, replace the directory with this file path to get detailed evaluation results.
