Datasets:

Modalities:
Tabular
Text
Formats:
csv
Size:
< 1K
Libraries:
Datasets
pandas
License:

What is passed to the LLM Equality Checker

#9
by sophies-cb - opened

Can you please clarify for the LLM Equality Checker when you say you pass the question that it does not include the data sources, it's is solely the question in the AA-LCR dataset "question" column and does not include any of the texts

Assess whether the following CANDIDATE ANSWER is CORRECT or INCORRECT.
For the CANDIDATE ANSWER to be correct, it must be consistent with the OFFICIAL ANSWER.

The question, for reference only: {question}
The OFFICIAL ANSWER: {official_answer}
CANDIDATE ANSWER TO ASSESS: {candidate_answer}

Reply only with CORRECT or INCORRECT.

Artificial Analysis org

Hi @sophies-cb -- can confirm that we only pass the LLM judge the question text, and not the document/context text

kartikey-aa changed discussion status to closed

Sign up or log in to comment