Add test split of the benchmark used in our paper.
#1
by
nonpirat
- opened
The columns are "question", "answer", "relevant_laws" and "reference_answer". "reference_answer" is the condensed version of "answer" without rationale and explanation. "relevant_laws" is in the form of list of dictionary [{"law": XXX, "sections": XXX}, ...]
LGTM!
Pawitsapak
changed pull request status to
merged