公开知识图谱记录

r-1

记录摘要

A single-pass document parsing model that integrates OCR, layout, table structur

从图片、扫描件或 PDF 中提取文字，并处理版面、表格或文档结构。

基准测试: hyperscaler products
基准测试: large LLMs
能力: text extraction from digital text, scans, handwriting
能力: table structure recognition including merged cells and nested headers
能力: layout detection including columns, headers, footers, reading order
能力: Integrates OCR, layout, tables, formatting, and grounding into a single full page pass
能力: Detects figures and provides a short generated description
部署方式: hosted Parse API V3

知识图谱当前记录了 13 个已知资料维度，另有 5 个维度尚未记录，关联 1 篇公开资料。这些数字只描述资料覆盖，不代表模型性能或适配性已经验证。