AI summaryⓘ
The authors found that even though document parsers seem to perform well on standard tests, they still struggle with complicated real-world tables. To better understand these problems, they created TableParseMap, a test set with tough tables showing different types of errors. They discovered that issues arise because big tables are hard to process at once, the visual clues are unclear, and the parsed results sometimes don't match the image. To fix this, the authors introduced DEC, a system that breaks down tables into parts, improves the visual info, and corrects mistakes without needing to retrain the parsers. DEC uses visual consistency checks to decide when to intervene and works well to improve parsing accuracy on difficult tables.
Table parsingTEDS scoreVisual consistencyTableParseMapDEC frameworkDocument understandingVision-language modelDiagnostic benchmarkStructural errorsAgentic system
Authors
Jutao Xiao, Yuan Qu, Dongsheng Ma, Fan Wu, Tianyao He, Weihong Li, Jie Yang, Yu Qiao, Bin Wang, Conghui He
Abstract
Recent document parsers achieve table TEDS scores above 93 on OmniDocBench v1.6, yet community feedback and our audit reveal persistent failures on complex real-world tables. To quantify this gap, we introduce TableParseMap, a diagnostic benchmark of 916 real-world tables organized into five challenging scenarios and nine failure types. The strongest evaluated parser achieves only 85.03 TEDS, showing that aggregate benchmark scores conceal substantial weaknesses. Our analysis attributes these failures to three complementary limitations: large tables exceed the reliable processing scale of a single pass, weak or ambiguous visual cues hinder structure perception, and the reconstructed table may remain visually inconsistent with the image. We therefore propose DEC (Decompose--Enhance--Correct), a visual-consistency-guided agentic framework that improves frozen table parsers without retraining. DEC uses a general VLM as the controller: Decompose partitions large tables along structure-aware boundaries, Enhance exposes weak visual evidence and reparses transformed views, and Correct diagnoses and repairs residual errors. A Visual Consistency Gate (VC-Gate) selectively triggers intervention, while a Visual Consistency Ranker (VC-Ranker) verifies candidate updates and supports rollback without ground-truth HTML at inference time. We further derive a 1,977-table Consensus-Hard Set from 4,556 candidates through offline metrics and cross-model consensus. Across three frozen parsers, DEC improves TEDS by 1.57 points on average; on TableParseMap, gains reach 1.89 points overall, 2.62 on structural errors, and 5.66 on large tables.