Updated NLI Verdicts to Binary Format #384
Conversation
this commit changes the verdict system in our natural language inference module from 'Yes/No' to a binary '1/0' format. This modification aims to streamline decision-making and ensure consistency across various languages, enhancing the system's efficiency and reliability in multilingual contexts
|
This is perfect timing @lucasiscovici , I was experimenting with the Hindi language and observed the same issue. I'll test this out. Do you have any ideas on similar issues/fixes in other prompts too? |
|
@shahules786 Great to hear that you've noticed the same issue with Hindi. This fix should definitely help with that. As for similar issues in other prompts, I'm looking into standardizing response formats across different languages to ensure uniformity and accuracy. If you've observed any specific inconsistencies or have suggestions for improvements in other prompts, I'd be really interested to hear about them so we can continue enhancing our system's performance. |
|
@lucasiscovici there are same issues in context_precision, recall, etc. Would love to include and test them in this PR. |
|
@shahules786 done |
|
Great @lucasiscovici , thank you. |
2 participants
this commit changes the verdict system in our natural language inference module from 'Yes/No' to a binary '1/0' format. This modification aims to streamline decision-making and ensure consistency across various languages, enhancing the system's efficiency and reliability in multilingual contexts