A Critical Evaluation of Evaluations for Long-Form Question Answering
May 2023
in “
arXiv (Cornell University)
”
Preprint — not peer reviewed. This was posted to a preprint server or data repository. It has not been through a journal's review process, and its findings may change or not hold up.
Studysummary This study found that current automatic text generation metrics do not predict human preference judgments for long-form question answering, and suggests adopting a multi-faceted evaluation approach focusing on aspects like factuality and completeness.
Automatically generated from the study's abstract, not written by a person, and not a review of the full paper. Not medical advice or a treatment recommendation. Read the original study, and consult a qualified healthcare professional before changing treatment. Full disclaimer
Read the full study on arxiv.org →