AI Paper Atlas

Research map / Large language models / Retrieval and knowledge

OCR Hinders RAG: Evaluating the Cascading Impact of OCR on Retrieval-Augmented Generation

Junyuan Zhang, Qintong Zhang, Bin Wang, Linke Ouyang, Zichen Wen, Ying Li, Ka-Ho Chow, Conghui He, Wentao Zhang

ICCV 2025 · Large language models › Retrieval and knowledge · approach: RAG robustness and evaluation · 64 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use a VLM (Qwen2.5-VL-72B) for OCR instead of pipeline MinerUOCR for RAG knowledge base · OHRBench overallF1: 30.0 -> 31.1Improved
Use best OCR instead of ground truth structured dataRAG system · OHRBench overallF1: 36.1 -> 31.1Hurt
Use VLM OCR instead of pipeline OCR for reading-order questionsQwen2.5-VL-72B · OHRBench reading order (overall)F1: 17.8 -> 8.8Hurt
Increase semantic noise from mild to severe (ablation)RAG retrievers and LLMs · OHRBench perturbed dataperformance: nearly 50% declineHurt
Add severe formatting noise (ablation)BGE-M3 retrieval and Llama3.1-8B generation · OHRBench multimodal questionsperformance: 12.7% and 9.1%Hurt
Use larger LLM (Qwen2-72B) under formatting noise (ablation)generator · OHRBench multimodal questionsperformance: 7% reductionImproved
Use HTML instead of Markdown/LaTeX table format (ablation)retrieval · OHRBench table questionsretrievalHurt
Use Markdown tables instead of HTML/LaTeX for generation (ablation)LLM generation · OHRBench table questionsgenerationHurt
Replace Llama3.1-8B generator with Ext2Gen-8B-R2 (ablation)RAG generator · OHRBench overall (Azure OCR)F1: 24.0 -> 42.8Improved
Use detailed instead of simple OCR prompt (ablation)VLM OCR · OHRBenchOCR/RAG performanceImproved
Rewrite questions by hand (Sim2Real) (ablation)MinerU · OHRBench 100 Q&AsF1: 20.7 -> 18.0Hurt
Use Qwen2-72B instead of Qwen2-7B as generator (ablation)error analysis with formatting noise · OHRBenchshare of OCR-error but correct answers: 9.4% and 3.5% higherImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy