This repository implements the 3-stage experiment described in the paper.
/risk
: Tests the baseline PT parameters/marker
: Implements experiment for marker mappings/risk_marker
: Replaces numeric probabilities with epistemic markers and runs PT parameter measurement again
pip install -r requirements.txt
Navigate to the directory corresponding to the stage you want to run. For example, to run baseline PT measurement:
cd /risk
Each stage directory has the following structure:
risk/
|---plot/
|---processed/
|---result/
analyze.py
elicitation.py
mle.py
prompt.py
values_probs.py
- Open
elicitation.py
- Go to line 256
- Select your model to test
- If your model requires an access token:
Then input your token when prompted
huggingface-cli login
python elicitation.py
Results will be saved under /result
python process.py > result.txt
- Measured parameters will be saved in
/processed
- Results will also be shown in
result.txt
To customize marker substitution:
- Open
/risk_marker/prompt.py
- Go to line 72
- Modify the function
safe_sub
as needed