UniRec-0.1B: Unified Text and Formula Recognition with 0.1B Parameters
2025/12/24 by Yongkun Du, Zhineng Chen, Du, Yongkun +10
Computer Science · #Computer Vision and Pattern Recognition (cs.CV) #FOS: Computer and information sciences #Handwritten Text Recognition Techniques #Natural Language Processing Techniques #Topic Modeling
paper · doi:10.48550/arxiv.2512.21095
openalex publication_date 2025/12/24 · openalex created_date 2025/12/26 · openalex updated_date 2026/07/28
Abstract
Text and formulas constitute the core informational components of many documents. Accurately and efficiently recognizing both is crucial for developing robust and generalizable document parsing systems. Recently, vision-language models (VLMs) have achieved impressive unified recognition of text and formulas. However, they are large-sized and computationally demanding, restricting their usage in many applications. In this paper, we propose UniRec-0.1B, a unified recognition model with only 0.1B parameters. It is capable of performing text and formula recognition at multiple levels, including characters, words, lines, paragraphs, and documents. To implement this task, we first establish UniRec40M, a large-scale dataset comprises 40 million text, formula and mixed samples, enabling the training of a powerful yet lightweight model. Secondly, we identify two challenges when building such a lightweight but unified expert model. They are: structural variability across levels and semantic entanglement between textual and formulaic content. To tackle these, we introduce a hierarchical supervision training that explicitly guides structural comprehension, and a semantic-decoupled tokenizer that separates text and formula representations. Finally, we develop a comprehensive evaluation benchmark covering Chinese and English documents from multiple domains and with multiple levels. Experimental results on this and public benchmarks demonstrate that UniRec-0.1B outperforms both general-purpose VLMs and leading document parsing expert models, while achieving 2-9x speedup, validating its effectiveness and efficiency. Codebase and Dataset: https://github.com/Topdu/OpenOCR.
Citations
- DocPTBench: Benchmarking End-to-End Photographed Document Parsing and Translation
- DeepSeek-OCR: Contexts Optical Compression
- 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
- Kwai Keye-VL 1.5 Technical Report
- PaddleOCR 3.0 Technical Report
- MonkeyOCR: Document Parsing with a Structure-Recognition-Relation Triplet Paradigm
- Dolphin: Document Image Parsing via Heterogeneous Anchor Prompting
- Qwen3 Technical Report
- Seed1.5-VL Technical Report
- InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
- Kimi-VL Technical Report
- SmolDocling: An ultra-compact vision-language model for end-to-end multi-modal document conversion
- olmOCR: Unlocking Trillions of Tokens in PDFs with Vision Language Models
- Qwen2.5-VL Technical Report
- Ocean-OCR: Towards General OCR Application via a Vision-Language Model
- OmniDocBench: Benchmarking Diverse PDF Document Parsing with Comprehensive Annotations
- SVTRv2: CTC Beats Encoder-Decoder Models in Scene Text Recognition
- Document Parsing Unveiled: Techniques, Challenges, and Prospects for Structured Information Extraction
- MinerU: An Open-Source Solution for Precise Document Content Extraction
- VL-Reader: Vision and Language Reconstructor is an Effective Scene Text Recognizer
- General OCR Theory: Towards OCR-2.0 via a Unified End-to-end Model
- Decoder Pre-Training with only Text for Scene Text Recognition
- Out of Length Text Recognition with Sub-String Matching
- DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model
- UniMERNet: A Universal Network for Real-World Mathematical Expression Recognition
- TextMonkey: An OCR-Free Large Multimodal Model for Understanding Document
- Instruction-Guided Scene Text Recognition
- InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Tasks
- Multi-modal In-Context Learning Makes an Ego-evolving Scene Text Recognizer
- Nougat: Neural Optical Understanding for Academic Documents
- Context Perception Parallel Decoder for Scene Text Recognition
- Revisiting Scene Text Recognition: A Data Perspective
- GPT-4 Technical Report
- ABINet++: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Spotting
- Self-supervised Character-to-Character Distillation for Text Recognition
- Multi-Granularity Prediction for Scene Text Recognition
- Scene Text Recognition with Permuted Autoregressive Sequence Models
- SVTR: Scene Text Recognition with a Single Visual Model
- Towards End-to-End Unified Scene Text Detection and Layout Analysis
- Focal Modulation Networks
- Benchmarking Chinese Text Recognition: Datasets, Baselines, and an Empirical Study
- OCR-free Document Understanding Transformer
- CDistNet: Perceiving Multi-Domain Character Distance for Robust Text Recognition
- Read Like Humans: Autonomous, Bidirectional and Iterative Language Modeling for Scene Text Recognition
- RobustScanner: Dynamically Enhancing Positional Clues for Robust Text Recognition
- Towards Accurate Scene Text Recognition with Semantic Reasoning Networks
- GTC: Guided Training of CTC Towards Efficient and Accurate Scene Text Recognition
- ICDAR 2019 Competition on Large-scale Street View Text with Partial Labeling -- RRC-LSVT
- Show, Attend and Read: A Simple and Strong Baseline for Irregular Text Recognition
- NRTR: A No-Recurrence Sequence-to-Sequence Model For Scene Text Recognition
- Watch, attend and parse: An end-to-end neural network based approach to handwritten mathematical expression recognition
- Image-to-Markup Generation with Coarse-to-Fine Attention
- SGDR: Stochastic Gradient Descent with Warm Restarts
- Synthetic Data for Text Localisation in Natural Images
- An End-to-End Trainable Neural Network for Image-based Sequence Recognition and Its Application to Scene Text Recognition
- Synthetic Data and Artificial Neural Networks for Natural Scene Text\n Recognition
Related