Skip to content
Hot eventLive

SP-DocReader:面向文档OCR的差异感知自博弈框架

1 reports1 sources3 hr ago updated

Get the story

AI overview

2026年10月9日,arXiv Computer Vision(一手)报道了SP-DocReader框架。该框架面向文档OCR提出自博弈方案:先进行监督微调,再针对残余错误继续训练。其核心方法包括Reading Discrepancy Masking,用最长公共子序列对齐参考与生成token,并对未匹配位置打分;以及Focused Fidelity Loss,在未匹配真值位置施加负对数似然监督。训练仅针对OCR模块,骨干网络保持冻结。

Generated from reports · updated 3 hr ago

Timeline

Follow the coverage from different angles.

Oct 9, 2026
  1. arXiv · Computer Vision
    SP-DocReader:面向精准文档 OCR 的差分感知自博弈框架

    SP-DocReader 提出面向文档 OCR 的自博弈框架,在监督微调后针对残余错误训练,Reading Discrepancy Masking 用最长公共子序列对齐参考与生成 token 并对未匹配位置打分,Focused Fidelity Loss 在未匹配真值位置加负对数似然监督,仅训练 OCR 模块、骨干网络保持冻结。

Heat trend

Current heat 9·Comparable peak 10(Oct 9)·Comparable change over 24 hours –

02.557.510Oct9Oct9Oct9Oct9

The trend compares only the same participants observed continuously; its range may be smaller than the current heat count. Move or click on the chart to inspect hourly heat; use the left and right arrow keys to switch.