olmOCR 2: Unit Test Rewards for Document OCR

2025/10/22 by Jake Poznanski, Luca Soldaini, Poznanski, Jake +3 · 3 citations
Computer Science · #Handwritten Text Recognition Techniques #Mathematics, Computing, and Information Processing #Natural Language Processing Techniques

paper · pdf · doi:10.48550/arxiv.2510.19817

Abstract

We present olmOCR 2, the latest in our family of powerful OCR systems for converting digitized print documents, like PDFs, into clean, naturally ordered plain text. olmOCR 2 is powered by olmOCR-2-7B-1025, a specialized, 7B vision language model (VLM) trained using reinforcement learning with verifiable rewards (RLVR), where our rewards are a diverse set of binary unit tests. To scale unit test creation, we develop a pipeline for generating synthetic documents with diverse and challenging layouts, known ground-truth HTML source code, and extracted test cases. We show that RL training on these test cases results in state-of-the-art performance on olmOCR-Bench, our English-language OCR benchmark, with the largest improvements in math formula conversion, table parsing, and multi-column layouts compared to previous versions. We release our model, data and code under permissive open licenses.

Citations

• PaddleOCR-VL: Boosting Multilingual Document Parsing via a 0.9B Ultra-Compact Vision-Language Model
• MinerU2.5: A Decoupled Vision-Language Model for Efficient High-Resolution Document Parsing
• DianJin-OCR-R1: Enhancing OCR Capabilities via a Reasoning-and-Tool Interleaved Vision-Language Model
• DocR1: Evidence Page-Guided GRPO for Multi-Page Document Understanding
• Seeing is Believing? Mitigating OCR Hallucinations in Multimodal Large Language Models
• Infinity Parser: Layout Aware Reinforcement Learning for Scanned Document Parsing
• Infinity Parser: Layout Aware Reinforcement Learning for Scanned Document Parsing
• olmOCR: Unlocking Trillions of Tokens in PDFs with Vision Language Models
• Qwen2.5-VL Technical Report
• Tulu 3: Pushing Frontiers in Open Language Model Post-Training
• MinerU: An Open-Source Solution for Precise Document Content Extraction
• Qwen2-VL: Enhancing Vision-Language Model's Perception of the World at Any Resolution
• Image Over Text: Transforming Formula Recognition Evaluation with Character Detection Matching
• General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model
• BioXP-0.5B: Explainable Medical-AI via RL-GRPO
• Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time
• Merging Models with Fisher-Weighted Averaging

Cited by

• Benchmarking Document Parsers on Mathematical Formula Extraction from PDFs
• MonkeyOCR v1.5 Technical Report: Unlocking Robust Document Parsing for Complex Patterns
• Freeing the Law with LOCUS: A Local Ordinance Corpus for the United States

Related

• olmOCR: Unlocking Trillions of Tokens in PDFs with Vision Language Models
• CC-OCR: A Comprehensive and Challenging OCR Benchmark for Evaluating Large Multimodal Models in Literacy
• DocVLM: Make Your VLM an Efficient Reader
• General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model
• Logics-Parsing Technical Report