Vision-language models (VLMs) have achieved strong performance on OCR-based benchmarks and increasingly focused on text-rich understanding, but their robustness under controlled visual degradation remains insufficiently understood. This gap is critical for OCR reasoning, where visual corruption can induce OCR errors and structural distortions, thereby introducing uncertainty into the reasoning task.
arXiv:2609.00232v1 Announce Type: new Abstract: Multimodal Large Language Models (MLLMs) have achieved strong performance on OCR-centric document understanding and text-rich visual reasoning benchmar...
arXiv:2608.07742v2 Announce Type: replace Abstract: Visual-language models (VLMs) frequently struggle with robustness issues in real-world situations due to low- or varying-quality input images. In t...
arXiv:2608.30678v1 Announce Type: new Abstract: Text-rich image understanding requires multimodal large language models (MLLMs) to organize OCR (Optical Character Recognition)-grounded evidence acros...
OCR-EDR is a rendering‑aware framework that diagnoses and repairs OCR errors by jointly evaluating an OCR prediction, its editable form, and the rendered image of the source document. It localizes genuine mistakes while preserving valid or rendering‑equivalent predictions, then applies executable edits and iteratively reassesses with updated renderings. On the newly constructed OCRErrBench, the DocEDR model achieves 94.78% diagnostic accuracy and repairs 86.23% of errors, boosting formula metrics by up to 30.99 percentage points and improving CDM scores on several OCR systems.
OCR-EDR is a rendering‑aware framework that diagnoses and repairs OCR errors by comparing an editable OCR prediction with its rendered image. It jointly assesses consistency, localizes genuine errors, and applies executable edits, optionally requesting updated renderings for iterative reassessment. On the newly constructed OCRErrBench, the DocEDR model achieves 94.78% diagnostic accuracy and repairs 86.23% of erroneous inputs, improving formula metrics by up to 30.99 percentage points on benchmark datasets.
The paper introduces a controlled benchmark for evaluating large language models (LLMs) on key‑value pair extraction from documents with varying levels of OCR noise. It tests 136 configurations across five instruction‑tuned open‑weight LLMs, three datasets, and four text‑quality conditions, using deterministic decoding to generate 17,688 document‑level inferences. The study finds that clean‑text performance does not reliably predict real‑world robustness, model rankings can reverse under noisy conditions, and few‑shot demonstrations do not always improve accuracy, highlighting reliability risks in OCR‑to‑LLM pipelines.
arXiv:2605.27750v2 Announce Type: replace-cross Abstract: Recent work has shown that Vision-Language Models (VLMs) used for optical character recognition (OCR) can generate plausible but visually uns...
arXiv:2609.37712v1 Announce Type: cross Abstract: Optical Character Recognition (OCR) is evolving from plain-text transcription toward general visual intelligence, requiring models to recognize, loca...
arXiv:2608. 12898v1 Announce Type: cross Abstract: Document parsing aims to transform unstructured documents into structured and machine-readable representations.
arXiv:2606. 08063v1 Announce Type: cross Abstract: Multimodal Large Language Models (MLLMs) have demonstrated remarkable success in visual understanding, yet their performance degrades significantly under real-world visual corruptions.
arXiv:2607. 09068v1 Announce Type: cross Abstract: Recent advancements in LVLMs necessitate robust benchmarks for complex, visually grounded reasoning.
The paper evaluates eleven vision‑language models (VLMs) for extracting structured fields from business documents, focusing on robustness, cost, and governance rather than just accuracy. Using a held‑out set of 750 synthetic checks, the study finds that fine‑tuning open‑source VLMs on 3,000 samples yields an F1 score above 0.98, surpassing all zero‑shot commercial systems, while GPT‑5 tops the commercial group and Claude Sonnet 4.5 fails on date extraction. The authors also present a practitioner‑oriented selection framework that maps task profiles—such as quality, latency, governance, and volume—to recommended approaches via filtering and total‑cost minimization, demonstrated on a mid‑volume document‑extraction scenario.