-
Notifications
You must be signed in to change notification settings - Fork 377
Prompts for climate_fever
#111
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
climate_fever
Signed-off-by: Matteo Manica <[email protected]>
|
hey @VictorSanh what are the next steps, should I add some more details or are there any changes I should work on? |
VictorSanh
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Yes, i think you are right, can you open a PR on datasets to correct this? + I also see that claim_label can have 3 as a value (Disputed)... I don't see disputed being used in the individual evidences though.
This looks great thank you!
Could you add some more templates?
For instance, one could be:
Here's a statement and accompanying evidence. Does the evidence {{"supports"}}, {{"refutes"}}, or provide {{"not enough info"}} on climate change?
Statement: {{claim}}
Evidence: {{evidence[...]}}|||
your_label
Exactly it is not used for the evidence. I will also open the PR on |
- paired templates as suggested in the PR review. - a template considering al evidence statements. Signed-off-by: Matteo Manica <[email protected]>
Thanks again for the quick feedback @VictorSanh. I added some new templates and fixed the labeling:
Will now open the PR on |
|
@VictorSanh As promised here the PR in |
VictorSanh
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
merging now, please update the spreadsheet!
Created five template prompt to evaluate claim/evidence pairs from the dataset.
One main note.
Looking at the schema from HuggingFace datasets (https://huggingface.co/datasets/climate_fever#data-fields), the labels should be indexed as follows:
But after some inspection I realized that 0 is usually associated to
SUPPORTSand 1 is associated toREFUTES.The template I created reflects this.