luoojason's picture
Markdown modality bundle (MarkItDown + Andrew OCR), mirrors PDF eval_bundle format; 1350 cells
f06250e verified
Raw
History Blame Contribute Delete
652 Bytes
{
"cell_id": "q115_control_k0",
"doc_id": "2023.findings-emnlp.248.pdf",
"question": "In the paper main results table comparing Standard, Chain-of-Thought, and the proposed method across the five logical reasoning datasets, on which dataset does the proposed method (without self-refinement) achieve its highest accuracy when GPT-4 is the underlying LLM?",
"answer": "LogicalDeduction",
"answer_format": "Str",
"condition": "control",
"k": 0,
"source_position": "first",
"expected_route": "azure",
"bundle_filenames": [
"source.md"
],
"original_filenames": [
"2023.findings-emnlp.248.pdf"
],
"modality": "markdown"
}