Back to feed
arXiv cs.LG
arXiv cs.LG
7/21/2026
DocOCR-Eval: A Correction-Based Framework for OCR Tool Selection Without Ground Truth

DocOCR-Eval: A Correction-Based Framework for OCR Tool Selection Without Ground Truth

Short summary

DocOCR-Eval is an annotation-free framework for evaluating and selecting OCR tools without ground-truth labels. It uses a three-stage correction and ranking strategy across multiple MLLMs to approximate annotation-based rankings. Experiments show reliable OCR tool selection is achievable in realistic, label-limited settings across diverse document domains and languages.

  • •Proposes DocOCR-Eval, an annotation-free OCR evaluation framework using correction-based ranking
  • •Aggregating multiple MLLMs improves alignment with annotation-based tool orderings
  • •Demonstrates reliable tool selection in label-scarce real-world document collections

Generated with AI, which can make mistakes.

#research-breakthrough#ai-tools
Read full article at arXiv cs.LG

Is this a good recommendation for you?

Comments

Failed to load comments. Please try again.

Explore more

AutoRAG vs RAGBuilder vs Red Hat AutoRAG: Which RAG Pipeline Wins on YOUR Data (and Their Shared OCR Blind Spot)
Dev.to
New AI systems speed document
Dev.to
Baidu Releases Unlimited OCR, a 3B Model That Keeps the KV Cache Flat for Long-Document Parsing
MarkTechPost
Benchmarking Frontier LLMs on Arabic Cultural and Sociolinguistic Knowledge: A Cross-Evaluation Framework with Human SME Ground Truth
arXiv cs.CL