{ "metrics_doc": { "prose_recall": "truth 段落按 '. ' 切句，归一化后 ≥4 词的句子在输出中出现的比例", "reading_order": "每段前 6 词作锚点在输出中的位置，相邻段对位置严格递增的比例；锚点缺失计 0", "table_rows_intact": "truth 每行，存在某条输出行包含该行所有单元格字符串的比例", "table_header_intact": "leaf_header（无则 header）全部单元格出现在同一输出行：1 或 0；只测叶子单元格同行，不测跨列分组语义", "table_pipe_detected": "输出是否含 markdown 表格（| 开头结尾行且下一行为分隔行）：true/false", "pages_covered": "mixed 的 page_anchors 各自出现在输出中的比例；第三个锚点在扫描页，无 OCR 时上限 2/3", "scan_text_recall": "scan 文本按 prose_recall 口径的召回；无 OCR 预期为 0", "equation_recall": "expressions 经数学标记归一化（去 $ / \\[ \\] 定界、\\mathsf{} 等字体包裹、\\{ → {、^{\\wedge} → ^、\\_ → _、去全部空白）后子串命中的比例；对所有工具同一口径，衡量数学内容是否可恢复而非输出格式", "equation_recall_exact": "expressions 仅做大小写/空白归一化的纯文本精确子串命中比例（旧口径，保留作对照：输出 LaTeX 标记的路径在此列必然偏低）", "elapsed_s": "lab_run 实测耗时（秒），透传", "error": "lab_run 记录的异常文本，透传；无异常为 null" }, "tools": { "markitdown": { "version": "0.1.8" }, "pymupdf4llm": { "version": "1.28.2" }, "pdfplumber": { "version": "0.11.10" }, "markovo": { "version": "2026-09-19.v3-media-unified" } }, "python": "3.11.14", "date": "2026-10-06", "cells": [ { "tool": "markitdown", "sample": "equation", "kind": "equation", "prose_recall": null, "reading_order": null, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": 1.0, "equation_recall_exact": 1.0, "elapsed_s": 0.0439, "error": null }, { "tool": "markitdown", "sample": "mixed", "kind": "mixed", "prose_recall": 1.0, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 1.0, "table_pipe_detected": true, "pages_covered": 0.6667, "scan_text_recall": 0.0, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0329, "error": null }, { "tool": "markitdown", "sample": "prose-single", "kind": "prose", "prose_recall": 1.0, "reading_order": 1.0, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0246, "error": null }, { "tool": "markitdown", "sample": "rotated-header", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 0.0, "table_pipe_detected": true, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0118, "error": null }, { "tool": "markitdown", "sample": "scan-image", "kind": "scan", "prose_recall": null, "reading_order": null, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": 0.0, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.008, "error": null }, { "tool": "markitdown", "sample": "table-merged-header", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 0.0, "table_pipe_detected": true, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.023, "error": null }, { "tool": "markitdown", "sample": "table-simple", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 1.0, "table_pipe_detected": true, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0128, "error": null }, { "tool": "markitdown", "sample": "two-column", "kind": "columns", "prose_recall": 0.1111, "reading_order": 0.8, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.046, "error": null }, { "tool": "pymupdf4llm", "sample": "equation", "kind": "equation", "prose_recall": null, "reading_order": null, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": 0.75, "equation_recall_exact": 0.75, "elapsed_s": 0.2054, "error": null }, { "tool": "pymupdf4llm", "sample": "mixed", "kind": "mixed", "prose_recall": 1.0, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 1.0, "table_pipe_detected": true, "pages_covered": 0.6667, "scan_text_recall": 0.0, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.4484, "error": null }, { "tool": "pymupdf4llm", "sample": "prose-single", "kind": "prose", "prose_recall": 1.0, "reading_order": 1.0, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.1768, "error": null }, { "tool": "pymupdf4llm", "sample": "rotated-header", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 0.0, "table_pipe_detected": true, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.185, "error": null }, { "tool": "pymupdf4llm", "sample": "scan-image", "kind": "scan", "prose_recall": null, "reading_order": null, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": 0.0, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.2136, "error": null }, { "tool": "pymupdf4llm", "sample": "table-merged-header", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 0.0, "table_pipe_detected": true, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.2102, "error": null }, { "tool": "pymupdf4llm", "sample": "table-simple", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 1.0, "table_pipe_detected": true, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.1985, "error": null }, { "tool": "pymupdf4llm", "sample": "two-column", "kind": "columns", "prose_recall": 1.0, "reading_order": 1.0, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.178, "error": null }, { "tool": "pdfplumber", "sample": "equation", "kind": "equation", "prose_recall": null, "reading_order": null, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": 1.0, "equation_recall_exact": 1.0, "elapsed_s": 0.007, "error": null }, { "tool": "pdfplumber", "sample": "mixed", "kind": "mixed", "prose_recall": 1.0, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 1.0, "table_pipe_detected": false, "pages_covered": 0.6667, "scan_text_recall": 0.0, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0141, "error": null }, { "tool": "pdfplumber", "sample": "prose-single", "kind": "prose", "prose_recall": 1.0, "reading_order": 1.0, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0095, "error": null }, { "tool": "pdfplumber", "sample": "rotated-header", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 0.0, "table_pipe_detected": false, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0044, "error": null }, { "tool": "pdfplumber", "sample": "scan-image", "kind": "scan", "prose_recall": null, "reading_order": null, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": 0.0, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0006, "error": null }, { "tool": "pdfplumber", "sample": "table-merged-header", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 0.0, "table_pipe_detected": false, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0045, "error": null }, { "tool": "pdfplumber", "sample": "table-simple", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 1.0, "table_pipe_detected": false, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0046, "error": null }, { "tool": "pdfplumber", "sample": "two-column", "kind": "columns", "prose_recall": 0.1111, "reading_order": 0.8, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0181, "error": null }, { "tool": "markovo", "sample": "equation", "kind": "equation", "prose_recall": null, "reading_order": null, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": 0.5, "equation_recall_exact": 0.0, "elapsed_s": 0.0, "error": null }, { "tool": "markovo", "sample": "mixed", "kind": "mixed", "prose_recall": 1.0, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 1.0, "table_pipe_detected": false, "pages_covered": 1.0, "scan_text_recall": 1.0, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0, "error": null }, { "tool": "markovo", "sample": "prose-single", "kind": "prose", "prose_recall": 1.0, "reading_order": 1.0, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0, "error": null }, { "tool": "markovo", "sample": "rotated-header", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 1.0, "table_pipe_detected": false, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0, "error": null }, { "tool": "markovo", "sample": "scan-image", "kind": "scan", "prose_recall": null, "reading_order": null, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": 1.0, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0, "error": null }, { "tool": "markovo", "sample": "table-merged-header", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 1.0, "table_pipe_detected": false, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0, "error": null }, { "tool": "markovo", "sample": "table-simple", "kind": "table", "prose_recall": null, "reading_order": null, "table_rows_intact": 1.0, "table_header_intact": 1.0, "table_pipe_detected": false, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0, "error": null }, { "tool": "markovo", "sample": "two-column", "kind": "columns", "prose_recall": 1.0, "reading_order": 1.0, "table_rows_intact": null, "table_header_intact": null, "table_pipe_detected": null, "pages_covered": null, "scan_text_recall": null, "equation_recall": null, "equation_recall_exact": null, "elapsed_s": 0.0, "error": null } ] }