[Question]:Consultation on the 63.4% accuracy on DocBench

Do you need to ask a question?

• I have searched the existing question and discussions and this question is not already answered.
• I believe this is a legitimate question, not just a bug or feature request.

Your Question

Dear authors, I am writing to consult regarding the 63.4% accuracy reported on DocBench. The paper states that accuracy on DocBench can reach 63.4%. However, I selected 10 documents from DocBench and performed an end-to-end test; the accuracy was only 40.54% (15/37). Attached is my test result: "llm_judge_accuracy": 0.40540540540540543, "llm_judge_correct_total": "15/37" I believe this is a legitimate question, not just a bug or feature request. Could you please provide the exact test procedure so that I can reproduce your results? Thank you very much for your time.

Additional Context

No response