Document parsing turns document images into structured content, and it must work reliably across varied layouts and acquisition conditions. According to the paper WeVisDoc: From Coverage to Capability for Robust End-to-End Document Parsing, existing training corpora are skewed toward common document types and clean digital pages, and simply expanding coverage does not clarify how to fix a parser's remaining weaknesses.
To address this, the authors propose WeVisDoc, a two-stage, data-centric framework. Stage I widens semantic, structural, and appearance coverage using heterogeneous data and structure-preserving degradation synthesis. Stage II then employs a held-out probe to measure the Stage I parser's residual errors within fixed visual-structural clusters.
Those diagnostics drive targeted data construction and a reallocation of the target-token budget, allowing the second stage to focus on the parser's specific failure modes rather than only adding more data.
The paper reports that WeVisDoc-4B reaches an Overall score of 95.38 on OmniDocBench v1.6 and a mean Overall score of 75.54 across the three PureDocBench tracks, ranking first among the compared end-to-end parsers in all four settings. Compared with Stage I, Stage II improves Overall scores for the 2B and 4B models on both benchmarks, with larger gains on the degraded PureDocBench tracks, including a 4.03-point gain for the 4B model on the Real Degraded track.